Model count = response treatment response*treatment/link=log dist=poisson lrci type3 obstats Notice, that we can also specify the saturated model by explicitly entering the main effects in the model statement:

The '*' sign in the response*treatment, indicates that we want the interaction term and ALL lower order terms that is, the saturated model. Model count = response*treatment /link=log dist=poisson lrci type3 obstats In GENMOD: proc genmod data=ski order=data treatment|response specification adds an interaction term to the main effects model. The LOGLIN statement (line) now specifies the saturated model. In CATMOD: proc catmod data=ski order=data To fit the saturated models, we need to specify or add an interaction term in our code. Let's evaluate the CATMOD and GENMOD output for VitaminCLoglin.sas and R code for VitaminCLoglin.R.