![]() For the output, we have the same parts as we did for the independence model, except now we have more parameter estimates. In these two model specifications under GENMOD, we are fitting different numbers of parameters, and the output sections on LR Statistics For Type 3 Analysis will be different, but the estimates of odds and odds-ratios will be the same. 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. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |