I have a data set with 7 potential covariates. At the moment I am testing each one individually against the outcome variable. I am using AIC/BIC to see which covariate fits the model best and building from there. The problem that I am having is that most variables have some missing responses. I have created a new variable called notmissing which codes a row as notmissing=0 if all covariates have a value for that row or notmissing=1 for the rows where one or more covariates has a missing value. What I want to do is run the proc glimmix and include a statement that only includes rows where notmissing=0 so that I can use the AIC/BIC. This is the code I have written: proc glimmix data=cmr.u5mrfull method=mspl ; where=(notmissing=0); class country regionID; model u5mr = regionID year ltuberprev/ dist=normal link=id dfm=sat; random intercept year / subject=country type=chol; run; I have run it for each variable but the number of observation used is different in each analysis so this makes me think the code hasn't worked properly or there is something wrong with the notmissing variable but I am not sure which one is causing the problem ... View more