Methods to enable interactive access to and use of data were the most frequent notable practice described in the study. Altogether, twelve of the agencies reported one or more practices that provide users the opportunity to work with information on the site beyond simply accessing and downloading records or files. In all but one case, the medium of interaction was some form of Web-based tool or application. The exception was a BLS telephone-based voice response system that allows users to request specific data tables or other extracts from files to be faxed or sent out. Since the interviews were conducted, the BLS has created Web-based interactive tools to provide that service, as well as additional features similar to the others described here.

While the goal of interactivity for users was common across these organizations, the methods and styles of interaction varied considerably. The capabilities available in these interactive systems can be described in terms of three main types:

• ad hoc queries that enable users to extract information in structures and combinations that don’t exist in native form in the repository,
• visualization tools to present images based on processing and analysis of information from one or more sites,
• complex analyses of information extracted from one or more sites, with results presented to the user on line, and
• searching and indexing tools to support user exploration of the contents of one or more sites in ad hoc ways.

In some cases, the interactive capability offered by a repository included just one of these types, while others involved complex combinations. These interactive features were often described as representing the direction of planned future developments. In the time since the interviews were conducted, a quick survey of the Web sites of these organizations showed considerable expansion of the capabilities described in the interviews. Where appropriate, these newer capabilities are included in the details below.

The most advanced and complex capabilities we found combined tools for ad hoc queries, searching, data integration, and analysis. These are combined in a single tool set known as NESSTAR, developed at the University of Essex (UK) and Norwegian Social Science Data Services in Bergen. It is a combination of a browsing tool to locate data in a distributed data set and analytical tools to carry our simple analyses and download data to local files for further work. It works off a central server and uses standard syntax and data exchange tools (XML and DDI) to link across diverse systems. The central server maintains metadata about accessible data sets, filtering and authentication mechanisms, and the operational tools. The data reside in distributed organizations that collaborate in the overall system.⁶ NESSTAR is employed by both the UK Data Archive and The Zentral Archive. The NESSTAR system is used in the Zentral Archive’s Eurolab, described above.

Similar capabilities, though not in an integrated tool set, are offered by the NCES. That Center’s Web site provides 19 features that interact with one or more data sets. These features range from searches for data about individual schools and school districts to building a table from existing variables, to filtering through higher education data to find a set of institutions that are statistical peers within a chosen institution. In addition, NCES provides an online analysis engine (the Data Analysis System) to perform correlation analysis with selected variables. It also provides a mechanism to request more complex analyses by submitting requests online, with results returned by email. This last service requires registration, but involves no confidentiality control, since only unrestricted data are available. In addition, the NCES sponsors the International Archive of Educational Data, to support comparative and institutional research, which offers the Data Analysis System and similar interactive tools. The primary distinction between the NCES access and analysis tools, compared to NESSTAR, is the data source. NCES accesses its own and closely linked Federal sources; NESSTAR can access any data set linked to the server.

The ability to process ad hoc queries for specific data, and to create tabular output was reported for several other repositories. The FHWA, BLS, Annie E. Casey Foundation, and Urban Institute have Web-based tools on their sites that support that type of interaction. The same is true for the Census Bureau, though census data available by this means have been purged of elements that would violate confidentiality requirements. The Urban Institute system also has added analytical capability that allows the user to generate cross tabulations of data from some of its data sets. The Kids Counts repository developed by the Annie E. Casey Foundation provides for extraction of longitudinal data at the national, state, and county levels. The Web site provides for creating charts showing trends in these data over multi-year periods (most for 1993-2000), and comparisons across several localities (e.g., comparing two or more states or counties with those states). This repository also provides color-coded maps of the US or counties within states, showing values for indicators chosen from the database (e.g., percentage of children living in poverty).

Ad hoc query capabilities to generate some kind of visual display were reported in government repositories as well. Map referencing to data through a geographic information system (GIS) was described by the US Census Bureau, NCES, and the Minnesota Data Center. The Minnesota Center provided users with CD’s that combined data and map-based interactive display functions. That Center’s Web site also provides static maps displaying various economic and social data does not provide interactive mapping capabilities. The NCES Web site allows searching for some data and reports through map references, in particular with the Data from the National Assessment of Educational Progress. The Census Web site, by contrast, offers a very high level of interaction with map referenced data, linking a sophisticated GIS with Census and other data files. That site allows users to select from a range of variables and display their distribution on highly detailed maps, with resolution down to the local political unit and census tract level. Users can change the resolution of the mapping and the variables mapped, and overlay combinations of some variables. This level of interactivity in a GIS display was the most advanced of those reported.

Other forms of visual displays were available to a limited degree. When extracting data about individual school districts, the NCES Web site displayed pie and bar charts of selected variables. The displays were not interactive to the extent of changing the content or display type, only selecting individual institutions to examine. Scatter plots of crime statistics are available from the Urban Institute’s Federal Justice Statistics Research Center site. That site plots frequencies of the crimes and prosecution by crime type for several years and geographic levels. The user can choose the data to be displayed, but not the characteristics of the chart. The NESSTAR tool used by the UK Data Archive and Zentral Archive also has the capability to generate charts from analysis of data extracted through that system. The user can choose from a menu of chart formats to generate a display.

Some type of searching capability was reported for all the Web sites with interactive features. Key word searches were common, providing search engine-type access to material on the site. A geographically referenced gazetteer was reported by the UK data Archive, through which a user could search for data availability by map location. The NCES site provides a more structured search facility, with options to narrow searches by parameters matched to the contents of NCES databases. Identifying desired data sources through metadata files was the search strategy used in the NESSTAR system and the ILSIS, developed at the Zentral Archive. Metadata-based search capability has the virtue of creating a type of virtual catalog of data sources according to the search parameters established by the user. A related method for providing access to data resources via user searches was under development by the BLS. They were exploring automatic tagging of text and other content to facilitate indexing and efficient searching. However, results of this effort were not available for this report.

There has been significant development in the availability and power of Web-based interactive tools since these interviews were completed. The Web sites of all the participating organizations have been expanded and new features added since then. It was clear from their plans described in the interviews and the evidence of recent development completed that Web interactivity is a high priority. In describing these plans, many of the respondents made clear the reasons for this priority. One was the desire to provide users with easier, more efficient access and enhanced analytical power. The other was the potential for increased efficiency and cost savings for the repositories by automating access and analyses, with the user directing the processes. By investing in user-guided or controlled access and analyses, the organization could provide the same or enhanced services at lower costs to their budgets. Most of the respondents mentioned budget pressures as a constraint on responding to increased user demands in any other way. Given this combined incentive for interactive functionality, development along that path is very likely to continue.