Logit Model (LOGLINEAR command)

  • To separate the independent (factor) variables from the dependent variables in a logit model, use the keyword BY. The categorical variables preceding BY are the dependent variables; the categorical variables following BY are the independent variables.
  • A total of 10 categorical variables can be specified. In most cases, one of them is dependent.
  • A DESIGN subcommand should be used to request the desired logit model.
  • LOGLINEAR displays an analysis of dispersion and two measures of association: entropy and concentration. These measures are discussed elsewhere 1 and can be used to quantify the magnitude of association among the variables. Both are proportional reduction in error measures. The entropy statistic is analogous to Theil’s entropy measure, while the concentration statistic is analogous to Goodman and Kruskal’s tau-b. Both statistics measure the strength of association between the dependent variable and the predictor variable set.
1 Haberman, S. J. 1982. Analysis of dispersion of multinomial responses. Journal of the American Statistical Association, 77, 568-580.