Have a large data set (n=7M) and am running hierarchical logit models with increasingly complex product terms. The four-way converges without problem. The three-way converges without problem. The two-way with one of the components of the interaction included as a continuous measure does not converge. The two way with the components as factors (i.e., categorical) converges without a problem. The main effects converges without a problem. I am just curious. Anybody every see this before or have a clue what could be going on? It is utterly baffling to me...
↧