| Previous: Contributors | Up: Bibliography | Next: Syntax |
Log-linear models are for modeling contingency tables, the cross-tabulation of discrete individual-level variables. Contingency table models take as the ``unit of analysis'' for the purpose of the statistical procedure, the cell of a contingency table. The ``dependent variable'' is then the count within each cell, and the explanatory variables indicate what categories the cells fall into. These models are highly efficient computationally since there are so few ``observations,'' but they are asymptotically equivalent to logistic regression models run on the unpacked individual level data.