Hi everyone,

I am analysing data from a cluster-randomized controlled trial that only had 24 clusters. The outcome is a binary variable and I am using a multilevel mixed-effects generalized linear model (using <meglm>) to determine the effect the interventions had on our outcome of interest.

Given the small number of clusters, we would like to use the Kenward & Roger degrees-of-freedom correction method to deal with the problem of inflated Type I error rate that small samples experience.

However, I have been unable to use the <dfmethod> option with <meglm> as it only seems to be possible with <mixed>. I am using STATA ver15.

I'd be grateful for any thoughts on how to work around this or if anyone knows if there actually is a possibility of running <dfmethod> with <meglm>.

Thanks,
Jamita