| Federal
Committee on Statistical
Methodology Office of Management and Budget |
FCSM
Home ^ Methodology Reports ^ |
Statistical Policy Working Paper 14 - Workshop on Statistical Uses of Microcomputers in Federal Agencies
Click HERE for graphic. MEMBERS OF THE FEDERAL COMMITTEE ON STATISTICAL METHODOLOGY (June 1986) Maria E.Gonzalez (Chair) Office of Management and Budget Barbara A. Bailar William E. Kibler Bureau of the Census National Agricultural Statistics Service Yvonne M. Bishop David Pierce Energy Information Federal Reserve Board Administration Edwin J. Coleman Thomas Plewes Bureau of Economic Analysis Bureau of Labor Statistics John E. Cremeans Jane Ross Office of Business Analysis Social Security Administration Zahava D. Doering Wesley L. Schaible Defense Manpower Data Center Bureau of Labor Statistics Daniel E. Garnick Fritz Scheuren Bureau of Economic Analysis Internal Revenue Service Terry Ireland Monroe G. Sirken National Security Agency National Center for Health Statistics Charles D. Jones Thomas G. Staples Bureau of the Census Social Security Administration Daniel Kasprzyk Robert D. Tortora Bureau of the Census National Agricultural Statistics Service PREFACE The Federal Committee on Statistical Methodology was organized by OMB in 1975 to investigate methodological issues in Federal statistics. Members of the committee, selected by OMB on the basis of their individual expertise and interest in statistical methods, serve in their personal capacity rather than as agency representatives. The committee conducts its work through subcommittees that are organized to study particular issues and that are open to any Federal employee who wishes to participate in the studies. Working papers are prepared by the subcommittee members and reflect only their individual and collective views. The Subcommittee on Statistical Uses of Microcomputers in Federal Agencies organized a one-day workshop held on April 24, 1985. This working paper is based on the workshop and discusses four topics: planning to buy and use microcomputers for statistical purposes; electronic data dissemination; applications of microcomputers; and expert systems. The report is intended to provide helpful guidance to Federal agencies in purchasing and using microcomputers for statistical purposes. The Subcommittee on Statistical Uses of Microcomputers in Federal Agencies was chaired by Terry Ireland of the National Security Agency, Department of Defense. MEMBERS OF THE SUBCOMMITTEE ON USES OF MICROCOMPUTERS IN FEDERAL AGENCIES Terry Ireland*, Chair National Security Agency Ken Berkman Michael Leszcz Bureau of Economic Analysis Internal Revenue Service Jay Casselberry Tom Nagle Energy Information Administration Internal Revenue Service Frederick J. Cavanaugh Ronald Steele Bureau of the Census National Agricultural Statistics Service Lawrence H. Cox Peter Stevens Bureau of the Census Bureau of Labor Statistics Richard Engels Linda Bouchard Taylor Bureau of the Census Internal Revenue Service Maria E. Gonzalez* (ex officio) Mark Winer Office of Management and Budget Office of Management and Budget *Member, Federal Committee on Statistical Methodology ACKNOWLEDGEMENTS The idea of a workshop as a focal point for proceedings on Statistical Uses of Microcomputers was suggested by Maria Gonzalez, Chairperson of the Federal Committee on Statistical Methodology. She also provided contacts in many Federal agencies, which made possible a broad Federal participation in the workshop. The planning of the workshop was done by the Subcommittee. Four topics were selected for the sessions of the workshop. The chairpersons designated by the Subcommittee organized each session. They were: Chairperson Session on Planning Lawrence Cox, Bureau of the Census Session on Electronic Data Ken Berkman, Dissemination Bureau of Economic,Analysis Session on Applications Ronald Steele, National Agricultural Statistics Service Session on Expert Systems Terry Ireland, National Security Agency The proceedings were prepared by the chairpersons and rapporteurs of each session based on input from the speakers. The Subcommittee thanks all the speakers in the workshop for their participation. Terry Ireland, who chaired the Subcommittee, and Norman Glick edited the final report. Linda Taylor ably handled all the organizational and administrative details of the workshop the real basis for a very smooth-running conference. -iii- FEDERAL COMMITTEE ON STATISTICAL METHODOLOGY WORKSHOP ON STATISTICAL USES OF MICROCOMPUTERS IN FEDERAL AGENCIES April 24, 1985 TABLE OF CONTENTS Page Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i Members of the Subcommittee on Statistical Uses of Microcomputers . . . . . . . . . . . . . . . . . . . . . . . . . . .ii Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . iii Introduction. MARIA E. GONZALEZ, Office of Management and Budget. . . . . . . . . . . . . . . . . . . . . . . . 1 Session on Planning. . . . . . . . . . . . . . . . . . . . . . . . . 3 Summary. Prepared by FREDERICK J. CAVANAUGH, Bureau of the Census. . . . . . . . . . . . . . . . . . . . . . 3 Introduction. LAWRENCE H. COX, Bureau of the Census. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 The Census Bureau Microcomputer Information Center. RONALD SWANK, Bureau of the Census . . . . . . . . . . 6 The National Security Agency Personal Computing Information Center. KATHY SCHNAUBELT, National Security Agency . . . . . . . . . . . . . . . . . . . . . . . .11 II Use of Microcomputer Technology at the Bureau of Labor Statistics. PETER STEVENS, Bureau of Labor Statistics. . . . . . . . . . . . . . . . . . . . . . . .13 Discussion. LAWRENCE H. COX, Bureau of the Census. . . . . . . . . . . . . . . . . . . . . . . . . . . . .23 Questions,and Answers . . . . . . . . . . . . . . . . . . . . .25 Session on Electronic Data Dissemination . . . . . . . . . . . . . .29 Summary. Prepared by JAY CASSELBERRY, Energy Information Agency. . . . . . . . . . . . . . . . . . . . . . .29 Use of Microcomputer Disks to Disseminate Information. STUART WEISMAN, National Technical Information Service . . . . . . . . . . . . . . . . .29 Cendata: Development and Implementation. BARBARA ALDRICH, Bureau of the Census . . . . . . . . . . . . .34 Electronic Dissemination of Perishable Information. ROXANNE-WILLIAMS, U.S. Department of Agriculture . . . . . . . . . . . . . . . . . . .38 Questions and Answers . . . . . . . . . . . . . . . . . . . . .40 Session on Applications. . . . . . . . . . . . . . . . . . . . . . .45 Summary. Prepared by THOMAS NAGLE, Internal Revenue Service . . . . . . . . . . . . . . . . . . . . . . . .45 -iv- Spreadsheets and Statistical/Econometric Applications in Econometric Research. LINDA P. ATKINSON, U.S. Department of Agriculture . . . . . . . . . .46 Spreadsheets and Data Base Applications Used by the Crop Reporting Board in Reviewing Survey Indications and Preparing Publications. GARY NELSON, U.S. Department of Agriculture. . . . . . . . . . . . .50 Manager's Perspective on the Acquisition and Use of Microcomputer-Based Graphics Packages. RICHARD W. HAYS, Internal Revenue Service . . . . . . . . . . .51 Current Applications of UNIX-Based Microcomputer Systems. BRIAN CARNEY, U.S. Department of Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . .54 Equipped for the Future? PAUL DOBBINS, U.S. Department of the Treasury. . . . . . . . . . . . . . . . . . .56 Concerns About Data Integrity, Security, and Accessibility in an Environment Where Microcomputers and Mainframes Are Interfaced. DICK SHIVELY, U.S. Department of Agriculture. . . . . . . . . .58 Questions and Answers. . . . . . . . . . . . . . . . . . . . . . . .61 Session on Expert Systems. . . . . . . . . . . . . . . . . . . . . .67 Summary. Prepared by NORMAN GLICK, National Security Agency . . . . . . . . . . . . . . . . . . . . . . . .67 Introduction. TERRY IRELAND, National Security Agency. . . . . . . . . . . . . . . . . . . . . . . . . . . . .69 Expert System Tutorial. GEORGE LAWTON, Army Research Institute. . . . . . . . . . . . . . . . . . . . . . .70 An Extension of Statistical Software to Expert Systems. JAMES J. FILLIBEN, National Bureau of Standards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .78 Editing and Imputation. BRIAN GREENBERG, Bureau of the Census . . . . . . . . . . . . . . . . . . . . . . . . .85 Discussion. MARK WINER, Office of Management and Budget. . . . . . . . . . . . . . . . . . . . . . . . . . . . .93 Questions and Answers . . . . . . . . . . . . . . . . . . . . .94 Appendix. Announcement of Workshop on Statistical Uses of Microcomputers in Federal Agencies. . . . . . . . . . .97 -v- INTRODUCTION Maria E. Gonzalez, Office of Management and Budget A subcommittee of the Federal Committee on Statistical Methodology organized a one-day workshop on statistical uses of microcomputers in federal agencies. The purpose of the workshop was to share information among federal agencies on the statistical uses of microcomputers. About 200 persons from federal agencies attended the workshop. The audience had an opportunity to ask questions and make comments in the discussion period of each session. All were acquainted with the uses of microcomputers. Some were also responsible for the planning of statistical uses of microcomputers in their agencies. The announcement of the workshop is included in the Appendix. Four topics were discussed at this workshop. 1. Planning of Statistical Uses of Microcomputers. The first session described three microcomputer information centers in federal agencies. The purpose of personal computer (PC) information centers is to familiarize the agency users with the PC potentialities. This session focused on planning#, implementation, and evaluation within federal agencies of statistical uses of microcomputers. The main questions asked were: Who should have microcomputers? For what purposes should microcomputers be used? In what configurations? At what costs? How will microcomputers coexist with central automatic data processing services? 2. Electronic Data Dissemination. This session dealt with different data dissemination methods. The discussion covered each agency's approach to data dissemination and the problems encountered in implementation. 3. Applications of Microcomputers. This panel discussion focused on the usefulness and weaknesses of microcomputer software and operating systems, the interface of mainframes and microcomputers, and factors affecting data integrity, security, and accessibility. 4. Expert Systems The methodological basis for expert systems was discussed and several examples were given. The examples describe current expert systems with statistical applications. The proceedings of this one-day workshop follow. For each session there is a summary, the presentations, and the discussions that followed. -1- SESSION ON PLANNING SESSION SUMMARY* The microcomputer technology of the 1980s is a personal and, therefore, a user-oriented technology. However, planning for microcomputer,technology is often very complex and causes many changes in the workplace. Program planners must take many factors,into account when planning the introduction of a microcomputer system into their organization. Three personal computer information centers were described: The Census Microcomputer Information Center of the Bureau of the Census The Personal Computer Information Center of the National Security Agency The microcomputer system of the Bureau of Labor Statistics The planning, management and evaluation of microcomputer technology at the Census Bureau officially began in 1983 with a meeting of the Executive Staff. Prior to that time, microcomputer technology testing and evaluation work was ongoing at the Census Bureau, but this was the first time that agency-wide distribution of microcomputers was discussed. The Census Microcomputer Information Center (CMIC) was established as a result of this meeting. To give greater emphasis to the importance of microcomputer technology, the Census Bureau located the Center in the Office of the Director with its manager reporting directly to the Associate Director for Administration. The purposes of CMIC are to assist employees in learning about microcomputer technology -- both from a user point of view and a manager/procurer point of view -- and to reduce the overall costs of microcomputer technology purchase and maintenance. Employees are.given access to various brands of hardware and software to test prior to purchasing. They are also given "hands-on" experience in the use of the newest in microcomputer hardware and software through special arrangements made with the various vendors and manufacturers. On-site training in the use of hardware and software is provided by,outside trainers, with the divisions paying the costs for their employees. Costs currently range from about $100 to $125 per person per day, which are quite favorable in comparison with commercial costs of similar training. The activities of the National Security Agency's Personal Computer Information Center (PCIC) started approximately 18 months ago, when NSA ---------------------- *Frederick J. Cavanaugh, Bureau of the Census. -3- established the PCIC to train employees in the use of PCs and vendor- developed software. It did not take long to discover problems of compatibility among various brands of microcomputers. Therefore, standards were established to ensure that: 1.All microcomputer systems at NSA are compatible with one another for effective communications and portability. 2.All systems are able to function using the UNIX operating system -- again, to allow for communications and portability. 3.The microcomputer systems-are supportable; that is, they must be easily and cheaply repaired. 4.The systems are secure, so as not to divulge secret information. NSA has set its microcomputer standards around the IBM PC and PC/XT in a UNIX-based environment (IBM's PC/IX) and its office automation standards around the Wang PC. BLS's microcomputer system is essential for efficient office operation, and BLS has kept this in mind in designing-and developing its system. The BLS Executive Staff is very supportive of the microcomputer system. In designing the microcomputer system at BLS, several critical needs have to be met. These include: 1.The need for a system that can readily provide terminal communication with mainframe computers. 2.The need for a system capable of communicating among various machines and those located in field offices as well. 3.The need to provide security for confidential information. BLS undertook research and experiments to determine which microcomputer system best met its needs. Upon completion of the research, a single system comprised of machines from a single manufacturer was implemented and a set of standards was developed around its operation and use. The present system includes over 100 IBM PC/XTs and three Ethernet (FIPS 1O7) local area networks. The microcomputer systems described in the presentations form a continuum from the experimental or user-oriented approach to the more standard production or program-oriented approach. However, despite a commonality of needs and objectives, each agency has chosen a different approach to planning and managing microcomputer technology. -4- INTRODUCTION Lawrence R. Cox, Bureau of the Census Welcome to the Workshop on Statistical Uses of Microcomputers in Federal Agencies, sponsored by the Federal Committee on Statistical Methodology. We begin with this session on planning. Microcomputer technology is the technology of the 1980's. It is a personal and, therefore, a user-oriented technology. However, its focus on the individual often can be misleading from a planning perspective -- at the agency or office level, planning and managing the use of microcomputer technology becomes very complex very fast. While encompassing important technical issues concerned with hardware, software and communications networks, this technology also quickly brings the planner face-to-face with the.business of managing and deriving improvements systematically from technological change. Inevitably, the introduction of microcomputers into an organization changes the workplace and the skills and orientation of workers. It presents new choices and often demands that these be made swiftly. In large organizations and small offices, the following.questions must be addressed: - where does microcomputer technology fit into the agency or office? -how should it be introduced? -how can the organization experiment and grow with this technology? -what must the organization do to plan and manage this technology effectively? - should standards be set for its use? which 'standards? how should they be set? by whom? how should they be enforced? - what sort of future decisions need to be made, and who should make them? We are fortunate today to have a panel-of experts in this field, whose experience should shed light on answers to these and other important questions facing the statistical program manager about to embark on the introduction of microcomputers into his or her organization. They speak-with the experience of individuals tasked with managing groups assigned these responsibilities in three different Federal agencies: the Bureau of the Census, the National Security Agency, and- the Bureau of Labor Statistics. -5- The speakers are: - Mr. Ronald Swank, Manager, Census Microcomputer Information Center, Bureau of the Census. - Ms. Kathy Schnaubelt, Chief, Information Resources And New Technology Branch, National Security Agency. and - Mr. Peter Stevens, Chief, Division of Communications and Computing Technology, Bureau of Labor Statistics. Until recently Kathy was Chief of NSA's Personal Computer Information Center and had direct responsibility for the functions we will discuss this morning. Ron and Peter have had these responsibilities on a continuing basis for some time. Each speaker will make a brief presentation on how the problem of planning the use of microcomputers was addressed in their agency. I will follow with a few comments by way of formal discussion, and we will then open the floor for discussion and questions from the audience. THE CENSUS BUREAU MICROCOMPUTER INFORMATION CENTER Ronald Swank, Bureau of the Census The words "microcomputer" and "personal computer" are often used in a manner that blurs their intended use. In the true sense of the word "microcomputer," the Bureau of the Census has been using microcomputers since 1968. The FOSDIC (Film Optical Sensing Device for Input to Computers) allowed us to film and input census forms to computers without manual data entry. In 1973 we attached IBM 6250 tape drives to Sperry mainframes, approximately three years.before Sperry announced similar availability, and experimented with the ATL automated tape library. In 1982, eight Apple II+ personal computers were used to do the Puerto Rican Economic and Agriculture Census data checking and editing. These projects established the feasibility of using microcomputers in much of the Bureaus work. The Bureaus organization (3500 employees at headquarters, 9000 nationwide) can best be described as 35 separate companies (divisions, in our parlance) sharing the same resources, computers, management services, etc. You can imagine the problem this presents in setting priorities, standards and general directives. All of the Census Bureaus funding does not come directly from Congressional appropriations. So there has been a great deal of discussion on:the best way of introducing microcomputer technology to the Census user community, funding it and not intimidating or alienating Bureau users. In 1983, a joint decision was made to establish the Census Microcomputer Information Center (CMIC). The Center with a staff of 4 was placed organizationally in the Director's Office for two reasons: (1) to show Executive staff support for technology and.encourage users to make -6- active efforts to become familiar with its capabilities and (2) to avoid turf battles. The CMIC is a clearinghouse of information for use by Census Bureau employees. The goals and objectives of the CMIC are to: - assist Bureau employees in their analysis of microcomputers; -provide access to and demonstrations of a variety of hardware, software and peripherals; -provide hands-on experience with microcomputers without capital investment by the individual divisions; -provide training on microcomputer hardware and software; -provide a clearinghouse for documentation, catalogs, and pointers to knowledge for microcomputers, end-user computing and office automation; - decrease the cost of hardware and software through more informed procurement decisions. With the direction of program managers, Census Bureau employees may visit the Center for information about microcomputers, for discussions of the characteristics of particular computers and the applicability of microcomputers to projects, or for hands-on experience on a variety of machines in an attempt to implement those projects. One can use the computers in the Center for weeks if necessary, experimenting with various software on different machines. The role of the Center is to help Census Bureau staff define their processing needs, advise them of applicable software and guide them towards suitable computer equipment. The CMIC contains the more popular microcomputers and the more popular software. Yet, there are significant numbers of microcomputers that may provide a unique perspective in the industry and may offer the best overall systems for a particular problem. Therefore, the CMIC also sponsors product demonstrations about those microcomputers that are not currently on display in the Center. CENTER OPERATION/USE The Center's hours of operation are 9:00 a.m. to 4:00 p.m. Census Bureau personnel may schedule time to use a particular machine, software package, tutorial or specialized peripheral device for one- hour segments. They may also request one of the Center support personnel to work with them. We generally have personnel in the Center from 7:00 a.m. to 5:30 p.m. Time before and after hours of operation is devoted to Center personnel, allowing us to gather and exchange information on the day's occurrences and to provide specialized support to executives. -7- Some of the typical questions arising on a given day may be as simple as: What's the difference between a hard disk and a floppy? How can I get specific information on a specific product and its capabilities? What kind of tutorials/training are available for Lotus, dbase, etc.? Why doesn't a package perform in a specific manner? WHAT's AVAILABLE IN THE CENTER The Center subscribes to approximately 40 periodicals dealing primarily with microcomputers and associated technology. About 20% of these magazines are provided free. Also the Center has a library of 300+ books dealing with microcomputers, hardware, software, peripheral devices, etc. These books are directed at all levels of personnel. The magazines and books are available for checkout by Census Bureau employees. The Center subscribes to Data Pro for microcomputer hardware and software. There are numerous other vendor- or industry-provided catalogues available for review in the Center: IBM Personal Computer and XT Software Guide The Blue Book for IBM Engineering and Scientific Progress The Book of Apple Software The Ratings Newsletter IBM Software Directory Many of the supply and peripheral device catalogues are provided by vendors Public domain software is available in the Center. Most of it was acquired from Capital-PC for IBM's and compatibles and the Freeloader 500 software for the Apple machine. This software is not copyrighted and is available for the cost of reproduction. We have found many useful utilities available that have saved our users much development time. Microcomputer software in the following categories is available: Communications software Mathematical Database management systems Specialized Electronic spreadsheets Statistical Integrated software Word processing Presentation graphics Utilities Programming languages This software is available for user evaluation. The end user determines whether the product will produce the required results. About 15% of our software was provided by vendors for use in the Center -- but only in the Center -- for evaluation purposes and not for production work. The Bureau's policy on copyrighted software is that it is not to be copied for any reason other than backup. -8- HARDWARE There are 2 IBM PC/XT's hooked to a local area network. A Sperry Model 50, a Wang PC a Grid, Apple Macintosh, peripheral devices, plotters,- printers, Polaroid palette, etc. are available. Many microcomputer vendors (43 to date) have come to the Census Bureau to demonstrate their products, and many have loaned their products for evaluation from 30 to 60 days, depending on product. Some of the vendors are: A & F Computers Sony Digital Equipment Fujitsu Olivetti Motorola Hewlett-Packard Exxon Data General Radio Shack ELECTRONIC BULLETIN BOARD In February 1985, an electronic bulletin board was placed into service to facilitate information interchange on product evaluations, user projects, etc. PROCUREMENT POLICY The Census Bureaus procurement policy evolved because of our organizational structure and our funding. While all procurement actions are to be processed and controlled through the Bureaus Procurement office, requests for ADP-related actions will continue to require some specialized processing. The justification and acquisition approval for microcomputer equipment and off-the-shelf software and supplies totaling less than $10,000 is delegated to the Associate Director level. The ADP staff no longer is required to review and approve such purchases. When the purchase order is sent to the Procurement Office, requests for sole source and brandname purchases costing more than $500 must include a brief justification. When the purchase order is received in the Procurement Office, information copies are forwarded to the Census Microcomputer Information Center to be used to update the Census Bureau inventory of microcomputer equipment and software. MICROCOMPUTER MAINTENANCE POLICY The Census Bureau's microcomputer maintenance policy is based on cost. For every six machines purchased we purchase a spare machine because the cost of a one-year maintenance contract on the first six equals the cost of the. spare. These machines are not just stored; they are used in noncritical environments where they can be removed to replace a critical machine as needed within one hour. When a user encounters an equipment problem that is beyond the users capability to resolve, he or she contacts, our Technical Services Division (TSD) service representative who will respond by sending a technician to the user's site to isolate the cause of the equipment problem. -9- If it is something simple that the technician can repair on the spot (such as replacing a fuse, resealing a loose board, or tightening a plug), the technician will make the repair. If the problem cannot be resolved by the technician on site, the technician will telephone the CMIC to request that a replacement computer or input/output device be loaned to the user until the user's machine is repaired. TSD will set up the replacement equipment for the user (if necessary) and take away the machine that needs repair. The user should be able to resume normal operations with minimal delay, aggravation and frustration. If the device is still covered by its original warranty, TSD will arrange to have it repaired under the terms of the guarantee. If the warranty is no longer valid, TSD will arrange to take the machine to a designated dealer for a repair estimate. When the machine is left with the dealer, a hand receipt will be signed by the dealer and returned to TSD. When the dealer calls the estimate to TSD, TSD will prepare a purchase request and forward it to the user's division. The division will insert the appropriate accounting code, approve the action, place a priority flag on it, and send it to the Procurement Office. The Procurement Office will expedite all micro maintenance requests by calling the dealer with a purchase order number. When the repairs have been finished and the machine is ready for pickup, a driver will take the purchase order to the dealer and pick up the machine. This procedure is valid for any repairs totaling less than $1,000. In cases involving repair estimates in excess of $1,000, TSD will contact the microcomputer user to discuss whether the repairs should be authorized and, if so, what procedure must be followed. The loaner machine will be under the control of the CMIC with the following priorities governing their use: Top priority -- to any user where TSD has removed a machine for authorized repairs. Second priority -- for use in support of hands-on training classes sponsored by CMIC. Third priority -- for use by someone who wants to do small projects on a borrowed machine. Priority will mean exactly that A broken machine will be replaced with a loaner from the CMIC even if it means having to take the loaner away from someone who is using it under a lower priority. I want to emphasize that this is our current policy, but it can be changed very quickly. We are constantly monitoring this procedure and continually reassessing our options (i.e., outside service contract). MICROCOMPUTER TRAINING SUPPORT We established a classroom with 16 machines for hands-on-training. We did this because of the numbers of people requiring training and the cost of sending people to outside courses. The types of courses taught are: Introduction to Microcomputers, Databases, Word processing, Spreadsheets, Graphics, etc. -10- Originally there were requests for training of 3000 persons in all aspects of microcomputers. That has been reduced to approximately 2100. We believe this training demand will be high initially and then will drop off dramatically. Outside instructors have been hired to teach our classes. We have had a great deal of success with this process because of the quality of instructors acquired. To pay for this training facility we charge back directly to the user division the cost of the instructor, software purchased and maintenance cost of the classroom. This cost goes to a maximum of $125 per class, significantly cheaper than to send all people to outside training. NOTE: Many vendors sell at a small cost educational licensing agreements providing copies of their software for each machine in the classroom. some vendors will not do this; then we must purchase copies for each machine at full price. OFFICE AUTOMATION I have specifically not addressed the topic of office automation, as we are still planning and discussing exactly what office automation is going to mean at the Census Bureau. Our primary planning focus at this time is to determine what functions need to be provided Bureau- wide and what functions will be left to individual operating units. THE NATIONAL SECURITY AGENCY PERSONAL COMPUTING INFORMATION CENTER Kathy Schnaubelt, National Security Agency The National Security Agency established a Personal Computing Information Center (or PCIC for short) approximately a year and a half ago. This action was taken in response to the Agency's growing demand for personal computer products. In the year prior to the opening of the PCIC, many new personal- computer products and vendors were reaching the marketplace. A growing number of these products were in turn being purchased by a cross-section of Agency elements. This mix of products across the Agency began surfacing problems such as that of system incompatibility. This may be illustrated by, the example of a diskette of data or software running on one computer brand but not on a different brand of computer. The PCIC was designed to assist Agency personnel in the selection, acquisition and use of an established set of "standard" personal computer products. The basis for the selection of standard products was determined by the Agency's needs as a whole. One such requirement was for the UNIX operating system. Hardware selected as the Agency standard workstation would have to be able to run under the UNIX operating system. At the root of decisions of this nature was the concept of compatible hardware and software products that would be easy for people to acquire. -11- Another important concern for us was security. By going to standardization, that problem may be minimized by the selection of products that meet this requirement and then training personnel to use them. A third consideration was supportability. Maintaining a variety of microcomputers, or personal computers, can be a logistics nightmare; stocking of parts, replacing them, etc., in any number can be devasting. Finally, there is cost. By limiting the number of kinds of personal computers and software products that we use, we are able to buy large numbers of each at a lower per-unit cost. Right now we have thousands of microcomputers in the Agency, and we have plans to buy many more, which should result in a significant savings from bulk buys. The PCIC was established to meet the following objectives: 1) to promote the use of standard equipment; 2) to share an centralize our small systems resources (like everyone here, we have a limited number of people to support these products); 3) to minimize the end-user application load; 4) to maximize cost effectiveness; and 5) to centralize product registration (providing anonymity in our workplace). The PCIC has become a focal point for all Agency standard products, and to date these products include: an Agency standard terminal/workstation which is an enhanced IBM XT; the standard office automation equipment which is the WANG Professional Computer; an interim standard local area network. So there will be a family of Agency standard host computers. The PCIC provides its customers with information on all of the standard products that are available; and this includes a reference collection of books, periodicals, in-house-developed working aids, research guides, comparison charts of the capabilities of the different products, and a referral service for technical questions. It also provides demonstrations of standard products. Anyone can go down to the PCIC and use one of the standard products, whether it's hardware or software. To encourage the use of the PCIC by Agency personnel, the PCIC tries to make the acquisition of standard commercial products as simple as possible. Rather than have each office go out and do their own purchase request, an authorized individual can come into the PCIC and request commercial software. The software is actually stocked in the PCIC. We have licensed some items (like CONDOR and MICROPRO products for example). By doing that, we have actually reduced some costs by 70%. Non-standard products may still be purchased, but on a limited basis. A non-standard product must be requested in writing. This request is reviewed by a software evaluation team to determine the validity of the purchase request. When a product offers a unique capability, it is purchased and evaluated. A favorable evaluation results in the product's being added to the list of standard products. A product. which does not offer any capabilities beyond the standard product line, or in fact is defective, would be placed on a prohibited- purchase list. In any case, the PCIC still does the actual purchasing, whether it's for a standard product or, an evaluation copy of a non-standard product. This saves the requester from, the paperwork of writing a purchase request document. -12- While the purpose of the PCIC is to furnish standard products, it also functions in identifying products that meet certain minimum requirements for Agency use. These products are added to the list of standard products to provide a flexible work environment for Agency personnel. The goal is not to restrict what people do or how they do it, but to make sure that the products they use are compatible with other products used throughout the Agency. USE OF MICROCOMPUTER TECHNOLOGY AT THE BUREAU OF LABOR STATISTICS Peter Stevens, Bureau of Labor Statistics I made the discovery when putting this talk together that I could take the various displays and shuffle them and present them in almost any order I chose. I not quite sure what the conclusion from that would be, but with this heady sense of freedom, I decided to start in the middle. Therefore, the first display you see discusses a brief introduction as to where we are now. The Bureau of Labor Statistics has approximately, 100 microcomputers, almost all of them standard IBM PC/XT's (see Display 1). We also have three Ethernet Local Area Networks two in D.C. and the other in the San Francisco regional office. We have network licenses and centralized software libraries for all of these machines. This is one point, and the first of the points which I will be emphasizing, where some of the things that we are doing that are, perhaps, different from what is commonly done. Floppy disks have no essential role in the entire operation. If I had my way, I wouldn't have them. Bureau of Labor Statistics Networks and Microcomputers Where we are now Approximately 100 microcomputers in use (mostly highly modified IBM PC/XT3) Three Ethernet (FIPS 107) Local Area Networks, two in DC, the other in San Francisco. Software libraries are centralized. We are close to completing the "large scale pilot" stage of our development effort. For each application area our goal is to identify and validate quality products which can be made a part of the standard BLS microcomputing environment. Display 1. In general, the way people get software onto their machines is through local area networks from centralized storage devices. We are getting to the end of what might be called the "research phase" of this entire new technology operation. The three networks were all acquired by a competitive -13- procurement which we ran a couple of years ago and which is, in effect, a large-scale+west. That gets to the last point on Display 1, which is the basic goal for what we are trying to accomplish right now: to identify and validate quality products which can be made A part of the standard BLS microcomputing environment; then, in the next stage of our operations, to make standards for use throughout the Bureau. When I looked at Display 2, I decided I could put it up and talk about it for twenty minutes without any trouble at all because it enumerates the applications and I think that gives some scope of the project. But given the terrible time constraints that we are under, I will spare you a lot of discussion here. The following are major application areas: Word Processing Graphics Spreadsheets Statistical Analysis Data Base Management Survey Data Collection Survey Control Project Management Calendar Management Network Services, including Electronic Mail, Shared Data Management and Inter-network Routing. National Communications via Public Value-Added Networks (X.25 & FIPS 100 standards). Mainframe Communications Gateways for Interactive and Batch Operations. Access to the Local Networks from remote (usually portable) microcomputers. Display 2. However, there are two things worth pointing out. Some may know from the previous references that "FIPS" stands for Federal Information Processing Standards, which are produced by NBS and which we are trying to follow. We have more standards than FIPS 100, and those things are, in general, a significant part of our operation. One other point, before moving on here, that I think is worth some mention: applications like word processing, graphics, and spreadsheets are stantard and well known; but the applications that I call here Survey Control, Project Management, and Calendar Management get into a function for the microcomputer which I don't think has gotten the emphasis it deserves. This is a Control and Management function. In the same sense tnat a microcomputer is a useful tool to use with a project management package, It is also used and useful for keeping track of one's personal calendar and the -14- ordinary flow of activities through the division. responding to technology, this is definitely a growing area. Anyway, enough for the present. The reason for Display 3 is not so much a chance to give you the details of how the Bureau operates, but to make a point that our efforts, in these areas were started in response to a serious and well-understood operational problem that we are having. The large, centralized mainframe computer provides, in our view, a very poor, very weak environment for the general area of interactive applications. How This All Got Started Throughout the 70's the Bureaus approach to computing relied upon two large, IBM-mainframe, computer centers accessed via dial-up telephone lines. While this environment served the large-scale, batch-oriented, survey processing well, other applications were served poorly: Interactive applications were very hard to develop, and response from the mainframe computers varied widely. Data communications were a constant source of problems, especially those with our Regional Offices. The proliferation of incompatible word processing equipment caused continuing operational problem and prevented any more ambitious office automation efforts. The most promising technical approach to solving these problems was: Powerful microcomputers for interactive processing. Local Area Networks for the heaviest communications and for configuration management. Internetwork and Mainframe Gateways for extended communications. Public Data Networks for national communications. Display 3. Again, I'm sure you wouldn't like to see me stand here and cry, so I'll spare you the details of the problems we have had with data communications since the AT&T; divestiture. The final point under the problem areas is again worth some emphasis. We have, I think, some thirteen odd different brands of word processors in place. None of them communicate with each other. This is a story that has -15- been, again, welltold. There was, in the Bureaus top management and operations management, a perception that this had caused us a great deal of difficulty and a very strong desire not to perpetuate that same sort of incompatibility and lack of communication in the new technology. The lower part of Display 3 shows briefly what we have selected as the technological underpinnings of the steps we are taking. Again, we could have, a long discussion on say, minicomputers versus microcomputers and the local area network services, but it is beyond the scope of this panel. will only mention that these issues were very seriously considered, and the choices listed were not made lightly. I would like to draw your attention to the phrase "configuration management." Having, let's say, several hundred microcomputers all using the same software packages would not be, in our view, sufficient to guarantee compatibility. Companies are constantly issuing new versions, and these new versions are frequently incompatible with each other. So you need not only to standardize with the level of machinery, but you need to do version control and configuration management to insure that the potential of a standard environment endures. One of the major functions of the local area network is that it makes it really possible to do this. If we wish to put up a new version of a particular procedure, we can do so. We can test it and then make that transition very easily. Back when I was planning this, I had visions of myself running down the hall with 500 floppy discs trying to distribute them. It was the horror of that nightmare that led us in that direction. Display 41 "How This All Got Started" is from a configurations perspective. I urge you not to take this too literally, but, in conjunction with Display 3, it does demonstrate the basic structure of the communications and technical environment. The large, vertical black bars indicate the local area networks themselves (that is, cable connections between machines in a single area). We use two computer centers: National Institute of Health and Optimum Systems, Inc. Those dotted lines indicate communications through the public telephone system. On the networks themselves we basically have two types of devices: The workstations (that is, machines that people use) and network services for file storage, printing, Communications, etc. Now we are at the point where we can get down to the most important part of this presentation. One of the things that I would like to try and share with you, from our experience, is an idea that I call, on Display 5, "Important Operating Assumptions." An assumption here means about the same thing that "theory" means in physics or chemistry. It means an idea that we believe and accept as true and act upon, but at the same time are constantly retesting and reevaluating. -16- Click HERE for graphic. Important Operating Assumptions No single supplier can come even close to supplying top-quality products for all our requirements. The best quality and most creative software development now is being done by independent (and frequently quite small) Software Vendors. Standards, de facto and formal, play a much more important role for the microcomputer market than they do for the mini or mainframe market. We can increase effectiveness and reduce risk by emphasizing, open systems and standards rather than by becoming locked in to one manufacturer's product line. The most reliable source of information about new products is our own testing. The selection, testing and integration of hardware and software are professionally very demanding tasks. Statisticians and economists should not have to become Microcomputer experts to use the equipment well. Quality in the initial selection of hardware and software is only the start of an effective operation. Support, maintenance, and especially release control for software are essential to long-term effectiveness. Planned and controlled redundancy is the best and, in many cases, the only way to achieve high reliability. Display 5. The first four items are a basic description of why we are interested in open systems" or open-systems interconnection. We have substantial experience with being in the tender and enveloping grasp of a single manufacturer and in discovering that manufacturer's products don't meet new needs, or that there is no way to interface some new piece of equipment to the existing equipment. THE MOST RELIABLE SOURCE OF INFORMATION ABOUT ANY PRODUCT IS OUR OWN TESTING. This point belongs in bold print because that is probably the essence of the whole project. The computer business has always been full of what I will call "hype": statements of doubtful truths, made just to sell equipment. The microcomputer business is, if anything, worse than the mainframe side of the business. We have found that things like articles and advertisements in magazines, the flowing promises of salesmen, and similar frivolities are simply not a basis upon which we can operate. We have certain responsibilities to our users in the, Bureau so that when we say something is going to work, they can expect that it will work. We can't then turn and -18- say that the salesman said it will work. Much of our validation is this testing of the product claims. The next two items on Display 5 deal with another very important aspect of our work. Doing the kind of validation that will cut through the hype is, in our view, a demanding task and not one which need be or should be placed upon the working statistician and economist. We have a very large number of users that want to use this technology. We have a much smaller number that wish to become microcomputer experts. We are trying to create an environment in which economists, statisticians, managers, clerical personnel, and the whole BLS community can use microcomputers effectively without having to go through the struggle and pain that is associated with selection, testing, and integration of the underlying technology. The last item in Display 5, I think, is very similar to the ones already expressed by Census. The way you get the reliability is through redundancy. One of the conclusions that followed from that idea is to use a standard configuration. Even though a particular machine may be intended for word processing and the machine next to it may be intended for statistical analysis, the underlying hardware will be the same. So that, if on the day the analysis is due, that particular machine decides to go out to lunch, the other machine can be used to finish the job. We are getting down toward the end, so we can summarize this by talking about the Project Goals and Current Policies (Display 6). You may remember that I mentioned there were three important problems that this research effort was attempting to address: the need to have an- environment in which we could create good interactive systems; the need to deal with our data communication flows; and a need to provide effective intercommunication between machines when used for statistical survey work, office automation, or any other purpose. Those were the goals and the motivation to start the project. They remain the goals. Every product we distribute must be thoroughly tested before full regional use. Some of the regional offices have very little background in data processing. What we put there had better work, because we don't have the travel budget to fix the mess if it doesn't. -19- Project Goals and Current Policies Project Goals: To solve the identified major problems with communications and interactive computing. To ensure that new products are thoroughly tested before being put into production system or into all Regional Offices. To open up new application areas, especially in the areas of end-u3er computing and office automation. To establish the basis for the continuing, orderly introduction of improved hardware and software. Current Policies: The selection, evaluation, procurement, and support of new products is centralized. Strong, de facto standards exist. The development of end-user applications is decentralized. The introduction of new products to Bureau production systems is closely managed. Pilot tests are required and high-level approval must be gained before production commitments are made. The emphasis on compatibility, full communications, and Bureau-wide usage is quite strong. Display 6. Finally, we see this whole technology as having opened up the potential to get into kinds of applications, that simply weren't being done at,all by any type of computer,,such as some of those personal and local organizational ones that I mentioned earlier. We now need to establish a basis so that we can continue to introduce, in an effective and orderly manner, new products and new technology that continue to pour out of the industry. From that, we have certain policies: the centralized selection, evaluation, procurement, and support of new products. There is some doubt as to whether we will be able to sustain a centralized procurement function because of some of the problems in government procurement which are beyond the scope of this presentation. In contrast to this centralization, the development of end-user applications is decentralized. That is, the way that persons use the machines for a particular personator organizational task is a matter of their judgment and their discretion. When we are talking about introducing this technology into Bureau production statistical systems, there is much stronger management control; and developments are closely watched. We insist on Bureau testing and evaluation before committing important Bureau projects to the new technology. -20- I think I have said enough about the need for compatibility. Finally, on Display 7, under the heading of Where We Are Going, there is basically more of the same. I mentioned we are getting toward the end of the large-scale research phase. We are planning to add local area networks into all eight regional offices instead of just San Francisco. We have one aspect of the Bureau which may be unique in that the Commissioner of Labor Statistics has A PC in her office. She also has one at home and uses them both. She has an intense personal interest in what I call here, "Management Communications." Through the local networks we have possibilities that we never had before. Through the research phase of this work, we have not had what I might call "traditional government procurement cost/benefit justification analysis" very much. I expect" as we move to the broader expansion of microcomputers into Bureau activities, that analyses of that nature will become important. There are many areas about procurement issues that are, at the moment, looking through a glass very darkly. Where We Are Going: As the performance of specific hardware and software products is validated, their use will be expanded to production tasks. The number of Local Area Networks will be expanded to include all Regional Offices. The communication facilities will be expanded to include Cooperating State Agencies for data collection and'survey processing. Management communications, among the Commissioner, Office Chiefs and Division Chiefs, will become increasingly important. The number of microcomputer workstations will be significantly expanded. Obsolete or ineffective equipment will be replaced by microcomputers. New hardware and software developments will be watched for possible replacements to standard products. As the new technology replaces existing equipment and applications, greater emphasis will be placed on cost/benefit justifications. Display 7. Display 8 shows where we expect to go technologically. I ask you not to take that too literally. This is not a technical model, but rather a demonstration of the way we see things getting done with each of the regions having its own network communicating to our network in Washington. -21- Click HERE for graphic. DISCUSSION Lawrence R. Cox, Bureau of the Census I will attempt to keep my comments brief so that we can have a full interchange between the speakers and the audience in proper "workshop- fashion. In proper "discussant" fashion, I will highlight what I see as the major similarities and differences among the three approaches taken, in the context of what I have learned from the presentations collectively and from my experiences at the Census Bureau. I have learned that microcomputer technology is a must for statistical programs. Automated, interlinked statistical program offices are more efficient and effective than those which are not. Users of statistical information have discovered microcomputer technology; and, so, statistical data providers have a responsibility to keep pace. Data review and analysis at its best is an interactive process between the expert data analyst and the data, supported by statistical software. Mainframe computing cannot offer these services on a large scale in a realistic manner or at a competitive price. I have learned that an organizational focus is needed to provide information and support both to management and users as this new technology becomes introduced and assimilated within the organization. We have seen that such a group can have any of several functions, depending upon organizational size,.needs, goals and objectives: -user education and handholding -repository of literature -source of hands-on experience -maintenance -training -develop and distribute product lists and recommendations -establish guidelines for microcomputer procurement, use, maintenance, training, etc. -recommend standards for microcomputer hardware, software and uses of microcomputer technology -establish and enforce such standards -aid in the procurement process -evaluate procurement requests -decide upon procurement requests -advise in the management of this new technology -play an active role in its management These functions, as I have presented them, lie on a continuum from the more passive, permissive or experimental approach to the more standardized, structured, or production-oriented approach. These needs and the management philosophies underlying them seem to me to be well-represented on that continuum by the three agencies represented here today. The free-market or laboratory approach adopted by the Census Bureau says, in effect, let's provide our diverse group of programs and users with the information necessary to begin to explore uses of microcomputer technology. Let's minimize the procurement obstacles to doing so, and let's work closely -23- with users in their applications and see what lessons are to be learned and what patterns emerge. In effect, as an organization, let's not force microcomputer hardware and software choices, but-let's closely manage and monitor several experiments and learn from each of them. At the National Security Agency, decisions were driven by the overriding need to standardize on hardware and software choices- sufficiently to allow diverse and distant groups to talk to each other and access the same data and programs, but stopped short of imposing inessential standards. Within a predefined architecture of standards, NSA users are free to experiment, to share information and to tailor choices to programmatic and individual needs. At the Bureau of Labor Statistics, the requirements for good and standard communications between offices and geographic areas were paramount. Experiments were conducted to fix upon the best choices, from which standards are to emerge. The environment is intended to be uniform and capable of supporting continuing, production-oriented work. Reflecting upon this continuum for a moment, I could equally describe it as being from user-oriented to program oriented, reflecting a progression defined in terms of the number of diverse programs and functions within these agencies which each agency seeks to address with automation at the microcomputer level. Interesting, all three organizations share several characteristics: they are not small, they deal routinely with massive amounts of data, their paramount concern is improved and broader access to their own data, their systems require mainframe-gateways or links, and they operate under strict data security requirements. However, for reasons which we have heard and others you may explore in open discussion, they have chosen three different approaches to tackling the problem of planning and managing microcomputer technology. QUESTIONS AND ANSWERS Ql: How was the Census Bureau able to acquire 500 microcomputers in a little over one year given GSA guidelines? Al (Mr. Swank): The Census Bureau did not go around GSA guidelines and standards, but worked within the existing regulations. Most procurements are off the GSA schedule. Q2: What variety does the Census Bureau have in their brands of microcomputers? A2 (Mr. Swank): Currently there are 25 different brands of microcomputers in operation at the Census Bureau. Q3: Does Census go through the "GSA microcomputer store" in procuring its microcomputers? -24- A3 (Mr. Swank): Yes, when possible. However, the GSA microcomputer store does not stock all brands, and this forces the Census Bureau to go elsewhere. Q4: Why did Census create a separate staff for microcomputers when they already had an established automatic data processing staff? A4 (Mr. Swank): The Executive Staff of the Census Bureau wanted to show support for microcomputer technology and to give it high visibility and, therefore, created the Census Microcomputer Information Center and placed it in the Director's Office. Q5: Has the Information Center taken an active role in education of upper-level management in the uses of microcomputer technology, A5 (Mr. Swank): Yes, each member of the Executive Staff has been given at least an introductory course on microcomputer usage. Q6: The presentation left several unanswered questions that should be addressed: 1. What about the lack of a management system for electronic files? 2. How are archiving and disposition of files handled? 3. What about programming for the PC's? A6 (Mr. Swank): Electronic filing systems will come in the near future. There are several such systems in existence now, but the costs are astronomical. A6 (Mr. Stevens): Software for record retention currently exists but the big problem is file retention for which very little software is available .Q7: Are the PC's at Census "stand-alone" or are they networked? A7 (Mr. Swank): Some PC's are networked others are "hardwired" to the mainframe; the majority are "stand-alones." Q8: Two questions regarding the presentations: 1. What is meant by "software standards"? 2. Some software packages need improvements, corrections, etc. In each agency, does anyone speak to the manufacturers as a representative of the agency? A8 (Mr. Stevens): "Software standards" means software standards. For example, there are at least three subcategories of word processing software, and each would have a separate software standard at BLS. A8 (Mr. Swank): Corporate licensing would be the answer. Those manufacturers that will not discuss corporate licensing have so much business they do not need to help and keep the client happy. -25- A8 (Ms. Schnaubelt): The focal point for NSA is with the vendor rather than the manufacturer. NSA has had problems with RUBIX from IBM. The smaller vendors are much more eager to get the business and give better contractual terms than the large firms. Q9: Is there a very strong recommendation from the panel for a PC information center? A9 (Dr. Cox) An independent PC information center is an absolute, necessity in a large organization. A9 (Mr. Swank) Each agency definitely needs at least a resource person if not a center. Q10 Would a small group need a PC information center? A10 (Dr. Cox): Not necessarily a center, but at least a reference person. Qll: Regarding machine-oriented versus people-oriented use of microcomputers, what would the individual agencies do for the people? What are the goals? All (Mr. Swank): At the Census Bureau, if the individual divisions have the budget, they will get the microcomputers they ordered within 3 0 days of the request. All (Ms. Schnaubelt): The goal is to have a PC on each desk. All-(Mr. Stevens): At BLS, the only drawbacks to a microcomputer on every desk are budget and procurement. Q12: With the advent of work-at-home, is there a use of portable PC's for this purpose? A12 (Dr. Cox): The major problem with portable PC's for take-home use is data security -- a large problem for each of the agencies represented. A12 (Ms. Schnaubelt): At NSA, portable microcomputers are used by executives and others, but these machines are kept "clean" (i.e., they have never had any sensitive data on them). The portables are used for training purposes only. A12 (Mr. Swank): The Census Bureau has many "checkout" machines, but some" of these are secure-machines and cannot be taken out of the building. A12 (Mr. Stevens): BLS definitely believes in the work-at-home concept and has machines for this purpose. However, precautions are taken to protect confidential data. Q13: How are services provided to field operators? A13 (Mr. Stevens): The regions do their own training on the uses of the BLS system. -26- A13 (Mr. Swank): There is a standardized configuration of microcomputer technology in each regional office with a nationwide company contracted to carry out maintenance. A13 (Ms. Schnaubelt): Data and software are transmitted world-wide by mail or other secured means of communication. -27- SESSION ON ELECTRONIC DATA DISSEMINATION SESSION SUMMARY The second session dealt with electronic data dissemination, focusing on disseminating information for use with microcomputers. While the first panel discussion focused on.how agencies use microcomputers within their own internal environments, this session deals with the impact of microcomputers on users of federal agencies' data and the possibilities for agencies to make information available for microcomputer users (that is, dissemination of data using floppy discs or through telecommunications). There are some very interesting opportunities for federal statistical agencies to use new media to provide data to users more quickly and in a form that is more highly usable than current printed methods. The three speakers will deal with these issues. The first speaker is from the National Technical Information Service (NTIS) which is primarily an archival-type agency for disseminating federal data and information. The NTIS program to disseminate data on floppy discs, the problems encountered, and the various issues surrounding this area will be discussed. The second speaker is with the Bureau of the Census and works with their telecommunications system called CENDATA. CENDATA is used to distribute perishable Census information to users. Our final speaker is from the Department of Agriculture. She will describe the current, ongoing process to implement a contract with the Martin Marietta Corporation to establish a telecommunications system for the dissemination of large databases containing agricultural information. USE OF MICROCOMPUTER DISKS TO DISSEMINATE INFORMATION Stuart Weisman, National Technical Information Service The history of the National Technical Information Service (NTIS) dates back to 1945 with the establishment of a publication board to assist in making unclassified government documents available to the private sector. The program went through various transformations, reaching its current status as an agency of the Department of Commerce in 1970. ---------- *Jay Casselberry, Energy Information Agency -29- The law creating NTIS states that NTIS is to search for, collect, classify, coordinate, integrate, record, catalog, and disseminate information. In the early 1970's, NTIS received its first machine- readable information product. In 1981 a new unit was established within NTIS to manage its product line of data base files and software. In the summer of 1984 NTIS began to sell data on floppy discs. The current NTIS machine-readable-products program contains about 10 bibliographic data bases, 300 source-text non-bibliographic data bases, 800 numeric and statistical data bases, and 1300 computer software programs. With this substantial amount of information available, NTIS began a review of procedures for disseminating information products for microcomputers. The following criteria were considered when NTIS reviewed the potential for disseminating their information products on microcomputer diskettes: -Forecasts of the number of microcomputers -Forecasts of,the primary type(s) of microcomputers being used by business and professionals -Physical size of the computer diskette -Microcomputer operating systems -In-house and/or contractor production of diskettes -Information products to be made available on diskettes -Entire and/or subsets of information files made available -Production of microcomputer software -Whether to reformat the data for use with popular data base spreadsheet formats NTIS has decided to make information products available on 5 1/4 inch diskettes for IBM and IBM-compatible microcomputers. Diskettes are produced by A contractor, and costs are determined based on the number of diskettes required. The main problems that have been encountered are in the loss or incorrect conversion of data when tapes or diskettes are produced, mishandling of diskettes during shipment, and improper use of the diskettes by customers. The way to overcome these problems is to establish procedures for checking a diskette against the original magnetic computer tape, and to instruct transportation companies and end-users on the proper handling of diskettes. In the future NTIS will consider producing information products on high density diskettes, hard discs, and, where it is practical, optical or video discs. With the future increases in microcomputers by business and professionals, NTIS is making a long-term commitment to having information products available for microcomputer users. With -the proliferation of data -30- management and analysis being done with microcomputers, NTIS recognizes the needs of this user community. Displays 9 through 16 illustrate the work of NTIS. HISTORY OF MACHINE-READABLE INFORMATION PRODUCTS Late 60's First machine-readable products arrive at NTIS Early 70's Production Group formed to process orders for machine- readable products Late 70's Concept of Product Management introduced 1981 Office of Data Base Services 1983 Video disc products from NASA 1984 Data files,available on diskette Display 9. DATA TAPES Over 1,000 Titles 32 Source Agencies 40 Titles Updated Annually 25 Titles Updated 2-6 Times a Year 15 Titles Updated monthly Remainder Updated Less than Annually Standing Orders Available Display 10. MAJOR DATA COLLECTIONS National Center for Health Statistics (NCHS) Federal Communications Commission (FCC) Energy Information Administration (ETA)/ U.S. Department of Energy National Bureau of Standards (NBS) Human Nutrition Information Service/ U.S. Department of Agriculture Defense Logistics Supply Center/ U.S. Department of Defense Federal Reserve Board (FRB) Environmental Protection Agency (EPA) Display 11. -31- DECISIONS, DECISIONS, DECISIONS!! Size: 5 1/4" vs. 8 1/2" (3 1/21, not readily available) Density: Double vs. single aided; Single vs. double-density (quad-density not readily available) MS-DOS vs. CP/M (or MS-DOS Vs. PC-DOS) Total in-house vs. contracting-out vs. in-house/out-house balance Products pre-selected vs. demand-driven selections Complete files only or subsets/extract3 Software ASCII only or various DBMS/spread3heet/fo ta Display 12. DATA DISKETTES 5 1/4" Diskettes Standard ASCII Fo-t For IBM-PC Microcomputer Unique Accession Numbers Assigned Data Tapes Converted to Diskettes Documentation Required Display 13. PLAYER RESPONSIBILITIES NTIS Contractor Source Agency Order Input & Control Create diskette master Provide master tape diskettes (with appropriate documentation) Copy tape to be used Archive Master Available for conversion consultation Ship Orders (with Duplicate Master documentation) Available for Get duplicates to NTIS consultation Available for consultation Display 14. -32- The Action Customer contacts NTIS "Available on Diskette?" YES NO 1. Price 1. Estimate price (based on #of diskettes) 2. Customer orders 2. Customer orders 3. Order to contractor 3. Copy master tape 4. Contractor duplicates master 4. Order to contractor with tape 5. Duplicate to NTIS 5. Contractor creates master diskette and duplicates master for customer order 6. NTIS mails (with documentation) 6. Duplicate to NTIS (price is to customer--overnight delivery actual # of diskettes) 7. NTIS mails (with documentation) to customer--overnight delivery Display 15. Problems Original tape ---------------- Bad tape from agency Copy tape at NTIS ---------------- NTIS error in copying tape Contractor converts ---------------- Contractor error in tape to diskette master conversion processor and duplicates master duplication process Duplicated diskettes ---------------- Problems created in sent to NTIS handling of diskettes NTIS ships diskettes ----------------(magnetic field, dropped, to customer smudge, coffee, etc.) Customer receives ----------------Customer mishandles diskettes and processes diskettes (see above) plus diskette processing Display 16. -33- CENDATA: DEVELOPMENT AND IMPLEMENTATION Barbara Aldrich, Bureau of the Census CENDATA is an information system for disseminating Bureau of the Census ("Census") information electronically. Development of CENDATA began in mid-1983 when Census decided that certain data, especially time-sensitive economic data, should be available on-line. CENDATA was developed under the guidelines that the data should be available on-line as soon as possible after release and that the system developed should be done at no cost to Census. The system was proposed as non-sole source, (i.e., not limited to only one contractor). In addition, no money was to be involved in the arrangement with any contractor, and Census was to have control over the information made available. During the entire process of developing the specifications @and establishing memoranda of understanding with qualified vendors, Department of Commerce lawyers assisted in refining the language and procedures. Census' list of qualifications for vendors wishing to access CENDATA and make the information available included: -A CENDATA user should only have to pay for time used accessing CENDATA -CENDATA should be available separate from other data bases, be clearly identified, and include the entire CENDATA package -CENDATA must be available seven days a week -A CENDATA vendor must be willing-to accept data delivery via telecommunications -A CENDATA vendor must be able to offer its users the services of national telecommunications networks -The system must be an end-use-based, user-friendly system The reasons behind the above qualifications were to: -ensure that vendors did not add hidden fees or package CENDATA with other services -enable users to use major telecommunications networks to minimize costs -obtain vendors with the capabilities to handle a large-scale data base such as CENDATA - increase dissemination of Census information products. Of the dozen vendors who have shown interest in the CENDATA system, four met the criteria established and; memoranda of understanding have been signed with two. -34- The first vendor, Dialog Information Services, went on-line with CENDATA on August 1, 1984. (Dialog is extremely prominent in the library community.) Dialog has CENDATA available using the standard menu-based system and also makes the information available in a full- text-searchable format. In-mid-October, 1984, the Glimpse Corporation made CENDATA available. Glimpse, in cooperation with the Chemical Bank of New York, markets data to the financial community. With the success achieved by the first two vendors in expanding the dissemination of Census data, Census is anticipating adding new vendors who service different sectors of the public. With the inherent advantages of CENDATA over traditional publications, Census hopes to continue to expand its user network. The primary advantages of CENDATA are the timeliness of the data and the ease of using the system. One of the first goals of CENDATA was to have sensitive economic information available within minutes after any embargo on the information is lifted. Examples of the- type of sensitive information. available are manufacturers' and shippers, orders, retail sales, housing starts, and balance of payments. Having this information available electronically assists users who are located away from Washington where the information is initially disseminated in press releases. The data are available weeks before users would receive it in published form, and it can be downloaded into a user's standard information system for review and analysis. Census also maintains an inventory of its products on CENDATA. This allows a user to quickly determine if a particular publication has been released, and, if so, the price, source, and Government Printing Office stock number. The illustrations that follow, Displays 17 through 21,.show how CENDATA has been developed for ease of use. Menus are designed to provide an inexperienced user with a choice of selections, and to move from general to the more specific. In addition, instructions are provided to help a user move through the system. THE CENDATA INTERACTIVE SYSTEM The Online Information Utility at the U.S. Census Bureau. A very small portion of the Census Bureaus vast data holdings has been included in this "information utility." Do you wish to see the CENDATA menu? If yes, enter Y or (return). If not, enter LOGOFF to end session. ?Y Display 17. -35- -- CENDATA MAIN MENUS 1 Introduction to Census Bureau Products and Services 2What's New in CENDATA 3U.S. Statistics at a Glance 4Press Releases 5Census User News 6Product Infoxmation 7CMMATA User Feedback 8 General Data 9 Agriculture Data 10 Business Data 11 Construction and Housing Data 12 Foreign Trade Data 13 Governments Data 14 International Data 15 Manufacturing Data 16 Population Data Enter item number or ? for help. ?15 Display 18. 15--MANUFACTURING 1 Introduction to the Manufacturing Statistics Program 2M3 Preliminary Report, July 1984 . . 8 Aluminum Ingot and Mill Products, June 1984 (CIR 1433-2) Enter item number or ? for help. ?2 Display 19. -36- 15.2--MX3 PRELIMINARY REPORT, JULY 1984 1 M3 Narrative Summary 2value of Manufacturers Shipments 3value of Manufacturers New Orders . . 7 Ratio of Manufacturers Inventories an Unfilled Orders to Shipments Enter item number or ? for help. ?3 Display 20. 15.2.3--August 30, 1984 TABLE 2, PART 1: VALUE OF MANUFACTURERS NEW ORDERS FOR INDUSTRY GROUPS, MARKET CATEGORIES, AND SUPPLEMENTARY SERIES --Seasonally adjusted- Monthly ( Millions of dollars) SIC Jul. Jun. May Code Industry 1984(p) 1984(r) 1984 All manufacturing industries. 192,450 190,620 193,680 Manufacturing industries with unfilled orders.............. 103,496 102,051 104,482 Durable goods industries............ 100,489 99,171 102,,256 --more- Display 21. After moving through the choices of information topics, the user is presented with the information requested. An experienced user may move through CENDATA more.quickly by specifying all parameters of its search at the same time. For example, by specifying 15.2.3 initially, all menus may be bypassed; and the user moves directly to manufacturing (15), the M-3 report (2), and specifically the value of manufacturers new orders (3). This development allows CENDATA to provide the necessary information and instructions for novice users without unduly hindering more experienced users. -37- As with any developing system, Census is soliciting comments from actual and potential users to determine possible system improvements and expansion of the data base. The primary users at the current time are economists,, industry analysts, and market researchers. Future plans are to, expand the data base with additional Census products. Upcoming products to be added are 1984 country population estimates and statistical profiles of every country in the world. With the addition of the statistical profiles, CENDATA moves into a new area since the information is from the International Data Base rather than from a publication, and the profiles are not readily available outside the system. ELECTRONIC DISSEMINATION OF PERISHABLE INFORMATION Roxanne Williams, Department of Agriculture The Department of Agriculture has as a primary function the dissemination of information about conditions related to Agriculture. The Extension Service is one way the Department uses to get information disseminated at the local level. In addition, the Department has long utilized the printed media for the dissemination of information around the nation. A few years ago, a number of agencies in the Department became dissatisfied with the print media because of the difficulty in getting information to interested parties as quickly as necessary. The agencies, acting independently, tried, electronic communication of data. Use was made of a number of commercial services such as 'DIALCOM, AGNET, and AGRADATA. DIALCOM is equivalent to an electronic bulletin board. AGNET is an on-line information system developed at the University of Nebraska. About two years ago, the Department started to have problems with the use of these services. Other information companies wanted the Department to provide the data going to existing services. They did not want to have to go to competitors for the information for a variety of reasons. One reason was that they wanted to be able to say they obtained the data directly from the USDA. Supplying each potential vendor with USDA data was just too much of a burden for the Department. In order to continue to get data to the ultimate end user and at the same time meet the needs of commercial vendors, it was decided to establish a single department-wide .system of electronic data dissemination. No agency will be forced to use this system; but if an agency decides to use electronic media, it must use the Department's system. This central system will then service the commercial vendors, including DIALCOM, AGNET, and AGRADATA. The Department decided to limit the scope of the project to what we call "time-sensitive perishable data." One example of this type of data is the agriculture marketing reports. These Are perishable because they contain the current prices and the current sales of all the different commodities around the country. The data are in constant demand and they are constantly changing as new reports arrive continuously. The demand for the quick and timely dissemination of these data is very high. -38- The Department is utilizing commercial vendor, Martin Marietta Corporation, to provide this service. This maintains a Department policy of not allowing public access to the Department's computer. It also keeps the Department from establishing a service that can be adequately provided by the private sector. Martin Marietta acts as an agent of the Department and has agreed not to use its position in order to benefit itself in the dissemination of these data to ultimate users. Martin Marietta can only disseminate these data through the system established for the Department. Other commercial vendors (we call them Level I users) can tie into the system with auto-dial or auto-set facilities. For a price, they can even have the main system's computer call their computer as soon as data ate released and transfer those data immediately. Thus all vendors will have excellent and "equal" access to USDA's perishable data. Equal access also meant to us that Martin Marietta would not charge other commercial vendors outrageous prices for access to the system. We wanted to keep the costs to Level I users reasonable. Martin Marietta was very reasonable and agreed to modest and uniform charges. Ease of access was also important to the Department. In order to maintain simplicity and keep programming costs low, we decided to use a straightforward file structure for the data with access obtained through a menu-driven system. The resulting simplicity of the system- not only makes -for easy access by users, but it also allows originating offices within the Department to upload files with a minimum of effort. Further, the originating offices maintain complete control over their own data in the system. They determine when data go into the system, when they are to be released, and when they are to be deleted. Martin Marietta only maintains the hardware and software of the system. In addition to meeting the requirements of outside (Level I) users, the system has been designed to the Department's own intern al requirements for information. A second type of user (Level II) has been defined. Level II users are primarily offices within the Department and the Extension Service. Other Federal agencies which make heavy use of agriculture data will be included. In order to service the Level II users, we asked Martin Marietta to allow access to smaller segments of data. These users do not need to obtain bulk data by telecommunications. The system allows us to break down bulk reports into smaller segments all of which are accessible via simple menus. The Department anticipates that the effect of the new system will be manifold. Users should have much better access to a wider range of information. Internal communication of information within the Department should improve significantly. The demand for hard copy should be significantly reduced. All of these effects should help to reduce the cost to the Department of data dissemination. -39- QUESTIONS AND ANSWERS Ql: What were the particular problems with mailing floppy discs; what kind of reject rates were encountered; and, if the discs are used for data transfer, how much of a backup do you need? Al (Mr. Weisman): Some problems in handling of the discs during shipment may have been avoided because we chose to use an overnight delivery service instead of the Postal Service. The quality of the service has been very high, there is very little handling required, and the service has not failed yet. Q2: Did you mention that there were some bad discs that needed to be replaced? A2 (Mr. Weisman): Yes. It is very difficult to track down where the mishandling of discs actually occurred. Q3: Is there a flat percentage of reliability? A3 (Mr. Weisman): The percentage of problems is very small, but it does occur. Q4: Has NTIS considered direct phone transmission of data; that is, could users call directly to the NTIS computer, similar to commercial data bases? A4 (Mr. Weisman): We did make our bibliographic data bases available similar towhat Census is now doing (as mentioned in the talk by Barbara Aldrich). That was started around 1974, or perhaps earlier. I believe there are now four vendors carrying our data base. In addition, NTIS encourages vendors to carry its statistical files and source files. To date, no vendor has elected to carry these files because it is more difficult to carry these files than a bibliographic data base. NTIS has no plans at this time to make these files available through telecommunications. Q5: What are the plans for disseminating data from the 1990 decennial census? A5 (Ms. Aldrich): In terms of data dissemination for 1990 decennial census data using CENDATA, there are no solid plans, but it is an issue for thought. The product information section of CENDATA could be used as a daily update or product release for 1990. I believe that there will be some electronic dissemination, but the amount and the level are not really being addressed at this time. Q6: Please tell us more about the software available through NTIS; is it public-domain software, software that the agencies have written for their own use, or some other type of software? A6 (Mr. Weisman): While I am the manager for data files and data bases and there is a separate product manager for software, I will try to answer your question. The criteria that NTIS uses for handling software are the same as, those used, for data files; that is, the software must be Government-produced. the software must also have a common usage and be useful to others. -40- Q7: NTIS currently sells a catalog of public domain software for $40 that includes quite a lot of information. Why doesn't NTIS publish separate catalogs of microcomputer software and mainframe software? A7 ( Mr. Weisman): At the present time NTIS only has three packages available on diskettes for microcomputers, the rest are for mainframes. NTIS does not convert software at the present time and may never do so. Currently there are not enough diskettes available for microcomputers to justify a separate catalog. Q8: Does Census have any feedback from CENDATA, users on the services and charges? AB (M. Aldrich): Yes, based on discussions with users, the charges seem reasonable. DIALOG priced CENDATA at $36 per hour, their most inexpensive commercial rate. That price does not include the telecommunications network charge which, with discount, is generally about $6 per hour. The Chemical Bank version of CENDATA is priced at $28 per hour and includes the telecommunications charge. In addition to the positive feedback we are receiving on prices, we receive feedback on what is in CENDATA, what users would like to see in CENDATA, and what they do not like. Q9: Is it possible to download CENDATA data and create other data files based on this? A9 (Ms. Aldrich): CENDATA is all public domain and no part is copyrighted. Therefore, it is available for users to download to their computers or add to other data bases. This caused a slight problem with DIALOG because so many of their data bases are copyrighted. To end any confusion, a notice was put in the DIALOG newsletter pointing out that CENDATA is in the public domain. Q10 (Mr. Berkman): Would Barbara and Roxanne discuss the impact upon their particular agencies' personnel who generate the data, in transferring the data to the two systems they discussed? A10 (Ms. Aldrich): I would like to cover the impact in two areas: the positives and the negatives. The negative for the people generating the data is that they must provide it to us in machine-readable form, either in the appropriate kind of floppy disc or via telecommunications to our microprocessor. There are some guidelines, with respect to designing tables that must be followed, which are quite difficult. The industry standard for CRT screens is 80 characters across, so any table must be defined in 75 characters since the vendors requested five characters for control. Often tables are split vertically, with the first part becoming Table 1, Part A; then the second part is Table 1, Part B; and so forth. The positive advantage to people preparing time-sensitive information and providing the data to CENDATA is a reduction in the interruptions from outside the agency with requests for data. Prior to CENDATA, when a data embargo was lifted, staff members would spend the remainder of the day answering the telephones and reading data over the phone. With the advent of CENDATA, users have an alternative where they can quickly receive the data. They can copy the data from CENDATA to their microcomputers and eliminate the need to listen to it over the phone and record it. There are both positives and negatives to the individuals who provide CENDATA with the information. In all cases the -41- individual division which is the source of the data provides the CENDATA staff with the information. A10 (Ms. Williams): Agriculture has designed a system whereby each agency retains control over its own data. This is a very sensitive subject, so the system was designed so that each agency enters its own data into the system. Because of the wide variety of equipment used to process data and create reports by our agencies, the system also needed to be designed so that the agencies did not need to change their current methods of doing business. To accommodate the agencies, each agency only needs to put a header card on its report to identify the report. If a report is to be broken up into different levels of service, an additional header card is necessary. Based on the header card(s), the system knows how to handle the report that follows. one agency, the Agriculture Marketing Service, required another accommodation because it used a leased wire service with a special protocol. Current users of these data had taps on the wire which were usually linked .to teletype machines. A microcomputer system was placed between their system and our system to convert the protocol and place the headers on the data. This allowed their system to operate exactly as it did prior to development of our system. Q11: Does CENDATA provide a computer tape to its vendors or is data communicated via telec ications? Also, how often are the vendors' files updated? All (Ms Aldrich): All CENDATA are transmitted via telecommunications. We use an enhanced word processor with telecommunications capabilities. Information initially goes into a private file where it is integrated into our standard system. We review the system exactly as a user would see it and,determine if there are any problems. Simple problems are corrected using the vendor's editor; serious problems may be corrected by deleting the file and starting over. When we give the go-ahead, the data become available on the vendors' systems. On DIALOG the files ate brought up overnight so the data becomes available the next day. We update daily based on data to be made available and changes in our product listings. The update is controlled by a vendor's software. We move records into and out of their systems. Q12: Does the Bureau of the Census pay for the update costs? A12 (Ms. Aldrich): No. Census developed the menu. We work closely with the software design people at each vendor. Q13: Do the vendors limit the amount of information? A14 (Ms. Aldrich) Certainly not in the case of DIALOG. They have the philosophy that however much information you can give them they will accept it. They consider data storage to be cheap and pride, themselves on being one of the largest vendors. In the case of Chemical Bank, they have not constrained us either. About once a year they request for planning purposes an estimate of how much storage we will need in the next two years. We have a small amount of data available on-line with a rich potential for it out of hand, but thus far there are no problems. -42- Q14: What were the reasons Census decided not to go sole source? A14 (Ms. Aldrich): One of the primary reasons was our objective to get the system operational as quickly as possible. By offering it to several vendors, we could avoid the procurement process. Another appeal was that by going with several vendors, CENDTA would be available to different segments of the community. With different vendors it might be possible to reach users that previously had not been Census data users. I think that in the case of DIALOG we have found a lot of librarians who were not previously users. Q15: Has meeting the different protocol requirements of the different vendors involved much extra work? A!5 (Ms. Aldrich): No, because we have only one system and one format for the data; each vendor must agree to adapt that format to whatever they see fit to use. There is one set of, codes which are very simple and straightforward. 43 - SESSION ON APPLICATIONS SESSION SUMMARY The relatively recent emergence of powerful microcomputers (micros) coupled with the availability of specialized vendor software packages for micros has significantly enhanced the federal statistical community's ability to gather, manipulate and analyze data. Today, more than ever, it has become easier to perform data analyses previously considered to be impractical due to resource and time limitations associated with traditional manual and computer methodologies. Accompanying enhanced analytical capabilities have improved methods for communicating the results of our data analyses. Powerful graphics software along with improved graphics plotters and color displays have made it possible to easily paint pictures reflecting data. analyses, which before were only possible through relatively expensive and involved mainframe processing. The boom in microcomputer usage in the areas of statistical and economic analyses is due in large part to the many advantages micros have over mini and mainframe computers. In particular, today's micros have storage capacities and processing speeds which often exceed mainframe capabilities commonly found just 10 years ago. Micros are generally simpler and easier to use than minis and mainframes; they are often portable; and they cost less to procure, operate and maintain. Micros are usually more reliable (less down time), and they often possess the ability to.communicate with minis and mainframes, which permits micros to access and transfer large data files. Along with the "hardware" advantages, there are also "software" advantages associated with micros. In particular, there is an abundance of high quality and user-friendly vendor software packages available, many of which permit the user to add his or her own code to modify and enhance the package's capabilities. Relative to mini and mainframe costs, these software packages are inexpensive. A few disadvantages of micros should be mentioned as well. The ability to exercise security measures and ensure control appear to be more limited. Today's micros are slow in comparison to current state- of-the-art mainframes. There exist serious compatibility problems of file structures between vendor software packages. Finally, there is often an added personal cost to the micro user in the area of additional time spent in procurement and maintenance, since these activities are usually not required of a mainframe user. The discussions which follow address many of the issues mentioned ---- above. ------------------ *Thomas Nagle, Internal Revenue Service -45- SPREADSHEET AND STATISTICAL/ECONOMETRIC APPLICATIONS IN ECONOMETRIC RESEARCH Linda P. Atkinson, U. S. Department of Agriculture Microcomputers are in widespread use throughout the Economic Research Service (ERS). I will be discussing their application not by secretarial staff for word processing or by data processing professionals, but rather by the economic research staff themselves. Our economists first became involved with microcomputers through the use of spreadsheet software, and this is still where the bulk of the applications are. Packages such as Supercalc and Lotus 1-2-3 are used extensively for data preparation, developing tabular reports, producing high-quality charts, graphs, and plots, performing if-then analyses, and interfacing with mainframe software. Some of the systems which have been developed with these packages are, in fact, quite sophisticated. One group, for example, has developed a program using Lotus 1-2-3 to assess preliminary economic impact of foreign pests to producers, consumers, and society in general. A partial budget analysis is used in which different economic scenarios are simulated by allowing changes in costs of production, yield, and prices for the affected crops. The entire system is menu driven and has options for various tables and graphs which can be produced. The program set-up is-being used as a template from which similar analyses can be developed, such as a program to evaluate the impact of change in ozone concentrations on yields. Another group hail been using Supercalc for data entry and preparatory calculations before running a program on the microcomputer to convert the data to-the form required for input to mainframe packages such as TROLL or SAS. After running these mainframe programs, files of output were then transmitted back to the microcomputer and reformulated for spreadsheet entry so that tables and graphs of output were automatically generated. Additional changes in the form of model output results could then be made, interfacing the flexibility of the microcomputer with the calculating power of the mainframe computer. Now this group has a simplified version of their model, the world grain-oilseeds-livestock (GOL) trade model, running entirely on the micro in Supercalc. The GOL model is an annual simulation model consisting of 27 country and regional models and 20 major agricultural commodities. The individual models are linked to solve simultaneously for a vector of prices which clear world trade. The global model system has equations for 339 country-commodity combinations. Running a 20-year projection on the full linked model on an IBM PC/XT took 48 hours; however, an individual country model runs in about 15 minutes. They hope to improve speed considerably by the acquisition of an IBM PC/AT with memory upgrades. The program has been. set up to ask questions of the user, such as what country is to be analyzed for what start and end dates. Users like the flexibility of the spreadsheet format; one can.get in and look at a simulation, watch the numbers change and see where any problems are. Built-in equation writers allow you to change the structure of a model or you can edit it directly. You can pre-create graphs and have them contain historical data to compare to simulated results. -46- A good reference on building such models in spreadsheets is an article from the February 1985 issue of Byte magazine entitled "Simultaneous Equations with Lotus 1-2-3." The author demonstrates how to formulate and solve a famous macroeconomic model, Klein's Model I, using standard Lotus commands. The Gauss-Seidel iterative method is used to numerically solve the system, with a one-line Lotus macro written to test for convergence. Another example of Supercalc use is to make projections of coarse grain production in foreign countries using population projections, real GNP growth rate, elasticities of consumption with respect to income, and growth rates of production. The spreadsheet format allows the analyst to change one item, such as an elasticity and have everything else recalculated. In this way it becomes easy to cross- check to see if implications of certain assumptions are reasonable. A planned enhancement to this analysis technique is to begin to use the regression capabilities of a microcomputer statistical package, ABSTAT specifically. Regression of grain conversions over time can yield estimated elasticities, which can then be put back into the spreadsheet. ABSTAT was acquired as a user-friendly package to do basic descriptive statistics and simple linear regressions. We have also acquired SPSS/PC, the micro version of the popular mainframe package. Many of our economists are accustomed to using SPSS for analyzing survey data and large cross-sectional data files such as those provided by the Census Bureau. To provide databanking of larger files of which portions might be analyzed using SPSS/PC, we recently licensed SPSS/X to run on our in-house minicomputer. SPSS/PC's ability to handle "portable" system files which can be uploaded and downloaded easily aids in forming an interface between the large and small computers. We will first apply this in analyzing the results of an in-house information-needs, survey; complete questionnaire results can be stored on the minicomputer, with data for particular groups of respondents or selected variables downloaded to the micro for detailed analysis without having to be redefined. We have two packages in-house that can perform more complex econometric estimation techniques: RATS (Regression Analysis of Time Series) and SORITEC. A domestic sugar model has been set up in SORITEC. Varioust estimations were performed, including OLS and two- stage least squares and Cochrane-Orcutt autocorrelation correction for each equation. The model was too large at 15 equations for SORITEC to do maximum-likelihood estimation of it, but the new version, when it comes, should be able to handle it. The model was simulated in SORITEC with the various sets of coefficients and also with various changes made to the model, for example perturbing an exogenous variable by 10% SORITEC has a command to compare actual and fitted values, computing summary statistics to measure goodness-of-fit. Because the model is somewhat large, it is run in a "batch" mode with Wordstar used to edit the SORITEC program. The model has also been put up on Lotus 1-2-3 to experiment with the parameters. Graphwriter is used to output plots of results. There is a free version of SORITEC called SORITEC Sampler which has capabilities of the main package up through, two-stage least squares. It cannot perform three-stage least squares maximum-likelihood estimation or -47- handle-nonlinear models. It produces nice screen graphics of regression plots including residuals, which can be dumped to a line printer (but not at present to a plotting device). While not of publication quality, the plots are very useful for analytical work. For example, as part of a farm production model, an equation was estimated with prices paid by farmers for feed as a function of corn price and the price,of soybean meal. The residuals showed some problems; an autocorrelation correction was tried and the regression re-estimated. The new plot showed substantial improvement in the residual analysis. Another analyst uses RATS to estimate import demand for wheat, corn and soybeans in four Asian countries. The 10-equation model has been run through OLS, instrumental variables and Taylor-series approximations, and he is trying to get around memory constraints (supposedly temporary until the new release of the package) to do seemingly unrelated regressions. The ARIMA time-series analysis capabilities of RATS were used in this project in determining how to average prices on a yearly basis" looking at the cross-covariances between prices and imports to decide on a lag structure. RATS is also being used to estimate a Canadian grains and rapeseed model. Again, a spreadsheet, in this case Lotus, is being used to update the data and provide graphical output, as well as to simulate the results. We have at ERS a number of other software packages for microcomputers to perform more specialized functions. GAUSS is a matrix programming language that allows you to write out an analysis the way you would write it mathematically. You can easily write down the estimation commands for the coefficients of a simple linear model, or the code for a complex statistical algorithm as it appears in a journal article. GAUSS does not currently come with built-in statistical routines but is planned to in the future. Another program, TK!Solver, solves simultaneous nonlinear systems, again allowing you to express the equations similarly to how you would mathematically. A package called MUMATH solves mathematical problems. symbolically land can take derivatives, etc. Especially useful in macroeconomic theory, one can change coefficients or other aspects of a model symbolically rather than numerically and see the logical implications in terms of cross-relationships that result. We even have some researchers who use small programs written in Basic to perform a specific statistical function, such as regression or the calculation of standard deviations or coefficients of variation, rather than bother learning how to use a more complete statistical package. Finally, I would like to mention one macroeconomic model to which ERS subscribes, FAIRMODEL which is a model-of the U.S. economy developed by Professor Ray Fair of Yale University-and programmed for the IBM PC and XT. The model consists of 30 stochastic equations and 98 identities and is re-estimated quarterly. It can be used for forecasting, policy analysis, scenario development and as a research tool. An analyst can run experiments change exogenous assumptions, enter adjustment factors, or exogenize an equation or block of equations, and view the results. An interface to Lotus 1-2-3 can be obtained with FAIRMODEL to use for setting up an analysis and deriving tables and graphs from the model output. --48- These have been only a few of the very many applications of microcomputers that we have in-house. The use of microcomputers has revolutionized the way our analysts conduct their research. In the area of econometric modeling, many more alternatives can be considered and assumptions tested in a much shorter period of time,, taking advantage of the interactive nature of the software oh these machines. Researchers who in some cases had little computer experience previously have become proficient with the easy-to-use and flexible software available on microcomputers, particularly spreadsheets, and seem to prefer this to the use of cumbersome statistical packages. However, now that better statistical software is becoming available, interest in it is growing. The economists I spoke with seemed to want to choose their, own components of an analysis system - spreadsheet, statistical program, graphics package, word processor - and are concerned with having good interfaces so they can quickly move data from one program to another. Some problems with memory constraints and speed have been experienced, but hardware is rapidly improving to alleviate this. There are worries about having errors creep