ABSTRACT
This chapter discusses the statistical models used to analyze the study data, and describe the information-theoretic approach to model selection that is the basis of that analysis. Extensive discussion of these models, and the information approach to model selection. It is convenient to start with the least squares regression model, detailing model selection for that familiar case. A common approach to this is to use hypothesis tests, such as N tests, to assess the significance of individual predictors, but this is actually not a very effective way of choosing the appropriate model to use, for several reasons. The chapter examines the question of estimating the probability that a trustee is black, using the various board characteristics as predictors. The disparity in black trusteeship between religious and nonreligious organizations is unusual in Los Angeles and Philadelphia, in Minneapolis/St. Paul, and, most strongly, in Atlanta.
