The Survey as a Basis for Symbolic Data Analysis

Survey data are initially presented as a micro file consisting of the necessary sampling design variables. Then the desired point-estimates as well as the interval estimates can be correctly computed. If several survey data files are available, and the results are needed for comparison, a usual way is to look at those estimates and to interpret the results in the best way possible. If there are a lot of estimates, the comparison may be hard since it is in fact needed to make a secondary analysis. If the analysis requires multivariate methods, the information loss is unavoidable. It is good if the loss is as minor as possible. Hence a new methodological technique, called symbolic data analysis, can be recommendable. It offers an appropriate tool to smartly aggregate survey micro data, and then to analyze such a symbolic file in order to summarize the basic estimates. This paper describes the key points of symbolic data and its analysis and gives a concretization using cross-national survey data, that is, the European Social Survey (see www.european.socialsurvey.org).