SIPP ACCESS represents a prototype for managing data of the social sciences and producing a scientific understanding of what is required to approach large-scale data bases. It is an experimental facility aimed at the discovery of improved techniques for creating scientific knowledge through the application of computer and information science to the social sciences. Its goals include (1) increasing the utilization of stored longitudinal data, (2) improving data retrieval from large-scale complex data collection activities, (3) clarifying the meaning of replicated measurements, and (4) training researchers in understanding such data in order to produce new knowledge of social and economic phenomena. The experiment incorporates an evaluation component to determine whether this conception of managing large-scale complex data can be appled to data other than SIPP (the Survey of Income and Program Participation). SIPP ACCESS mobilizes new technologies to make complex panel data available to researchers, analysts, and policymakers at low cost. SIPP ACCESS links measurements to documentation, online data to a national network of users, and results of research to users, thereby accelerating research developments. Its public utility aspect creates economies in data handling by reducing the complexity of materials that a researcher must deal with and by reorganizing replicated measurements into a form consistent with the conceptual structure of analytical models of dynamic phenomena. Efficiencies in retrieval and clarity in the longitudinal design of SIPP are achieved through eliminating redundant data and through restructuring cross-sectional data into time series and event histories. Data and documentation are integrated through an information management system that improves the delivery of data by eliminating delays, clarifying the meaning of measurements, and ensuring the retrieval of appropriate measures for analysis. Interactive access to data and documentation is provided through a national dial-in system, a relational data base management system which allows researchers to "build" their own data files and reorganize, transform, and link data over time according to their particular research questions. Optical archival data storage eliminates all programmer interventions between users and the data base. The data base management system improves retrieval of data from very large files, thereby assisting researchers in solving problems related to analysis such as identifying longitudinal samples, converting events to time series, time series to event histories, and identifying inconsistent data. Training workshops are used to introduce researchers and policy analysts to the subtleties of longitudinal data, relational theory, and the relational data base management system.
