The main objective of the study is to examine model selection methods in loglinear analysis. Log-linear analysis is a tool for independence analysis of qualitative data. Cell counts are Poisson distributed and all variables are treated as response. This method allows to analyze any number of variables in a multi-way contingency table. In log-linear analysis we model cell counts, where expected cell frequencies are functions of parameters representing characteristics of the categorical variables and their relationships with each other (interaction). The purpose of this paper is the presentation and comparison of model election criteria. The most popular statistics are chi-square test, likelihood ratio test and information criteria (AI...