Electronic statistics textbook banner

Statistical Advisor, Searching for Relationships/Patterns in Tables

Generalized Linear Models (GLZ)

Generalized Linear Models (GLZ) discusses an implementation of the generalized linear model and allows you to compute a standard, stepwise, or best subset multiple regression analysis with continuous as well as categorical predictors, and for binomial or multinomial dependent (response) variables (see Link Function).

Log Linear

Log Linear discusses various techniques for the analysis of multi-way frequency (crosstabulation) tables. Some software programs include an automatic model selection feature that will automatically select a best model for the data, that is, it will determine which factors in the table are related to each other.

Multi-way frequency tables arise when one crosstabulates multiple categorical variables. For example, one might study staff turnover (a yes/no categorical variable) as a function of occupational level, gender, ethnic background, etc. Based on those categorical variables, the program will compute a multi-way crosstabulation table.

The automatic model selection feature in LOG LINEAR will then suggest which variables (or factors) are related to turnover. Interactions between factors will also be revealed. For example, it is conceivable that only white Caucasian males in a particular occupational level are more likely to leave the organization (i.e., fall into the category turnover=yes). This example illustrates an interaction between the factors gender, ethnic background, occupational level, and turnover. In complex tables (e.g., 9-way crosstabulation tables), a very large number of such interactions may exist, and the program will automatically search for them.