hi all, Prior to this post I asked a question how to deal with binary clustered data when the crosstab by variables shows an empty cell. I figured out later that such an empty cell corresponds to what is called "separation". Fortunately, R package "logistf" is developed to handle such a problem using a penalised-likelihood method. However, it seems that logistf is only suitable when observations are completely independent, thus will be problematic if used on my clustered data. I was wondering if anyone had similar experience and has any suggestion for this? Thanks very much! Yue -- View this message in context: http://r.789695.n4.nabble.com/separation-occuring-in-clustered-data-tp4635770.html Sent from the R help mailing list archive at Nabble.com.