

Chisquare test for interaction effects 

Message/Author 

yao lu posted on Friday, March 16, 2012  2:36 pm



Hi Dr. Muthen, Sorry to trouble you again. I am using XWITH, Random, testing an interaction effect. In order to use Loglikelihood chisquare difference tests: 1. should I compare the interaction model with the linear model, or with the model assigning 0 to the interaction effect coefficient? 2. if it is comparing to the linear model, should H0 or H1 in the linear model output be used for log likelihood chisquare tests (i.e.,to be used with the H0 got from the interaction model)? Thank you very much. 


The z=test of the interaction is the same as doing a difference test or a Wald test. 

yao lu posted on Friday, March 16, 2012  3:12 pm



Thank you for the clarification. Just in case I need to conduct a difference test, can you still advise me on the two questions? To confirm, for Z test, it is under the column of "Est./S.E."? Thank you very much. 


You compare the model without the interaction to the model with the interaction. The only difference in the models is the interaction term. Yes, the ztest is the ratio of the parameter estimate to its standard error. 

Back to top 

