Читать книгу Elementary Regression Modeling - Roger A. Wojtkiewicz - Страница 21

In the discrete approach to regression analysis that I use in this book, individual units and groups of individual units have clearly defined roles. The key analytical unit in the discrete approach to regression analysis is the group. Many of the hypotheses that we test using regression analysis in the social sciences usually involve how one group is different than another group in some way. The discrete approach builds directly on the analysis of group differences as in the t test or cross-tabulations.

The role of the individual unit of analysis is that individual characteristics are summarized statistically and the summary statistic is used to describe the group.¹ The groups analyzed in regression analysis have no measured characteristics other than the summarized characteristics of the units in the groups.

For example, the mathematics ability of the group defined as students in private schools may be measured by the mean scores of students on a mathematics test. We do not measure directly the mathematical ability of the group but instead summarize the mathematics ability of students in the group and use that statistic to describe the group. The social science question addressed by statistical analysis is not whether individuals are different from one another but whether the groups that individuals belong to are different from one another.