hi, when I am estimating the model, it seems that one chain out of 4 would always fail. The message and the traceplot look like the following. I wonder if anyone could give me some general idea why this happens. This is very strange because the other 3 chain look fine and agree with each other. Is it because the algorithm stuck at somewhere?
It’s hard to see what’s going on form the trace plots you’ve provided. Is that blue line above 0 in the first image the trace of a chain? The usual reason for chains getting stuck is bad posterior geometry. The way to fix it is to usually use a better parameterization. The divergent transition report is also suggesting the model is struggling with the posterior geometry. Without seeing your model, it’s hard to say much more.