We propose a discrete random effects multinomial regression model to deal with estimation and inference issues in the case of categorical and hierarchical
But, looks like in the case of no adjustment for (categorical) confounders, the CMH test is not the correct one and multinomial logistic regression should be used. If so, how do I get a p-value for trend with 1 dependent and 1 independent variable as mentioned above. Any example cod...
Whilst I understand conceptually what this is, I have not found a way to do this in SAS yet. I don't know what you mean by "to do this in SAS." What is "this"? You must have used my article on regression splines to write your code. The article contains links to the SAS ...