I am performing a factor analysis on a set of binary variables. What I need from the model are the factor scores only, which I will be using for further analysis. My method requires that the factor scores are normally distributed. But in factor models with binary data that may not be always true. In fact, in my case the factor scores are very skewed. Are there any models for binary data that produce normal (or approximately normal) factor scores? Thanks.
Even if your prior (by model assumptions) is normal, the posterior that the factor scores come from is not necessarily normal. The posterior incorporates information from the data, so I don't think you want to force normality. Instead, change your method that you mention so it doesn't require normality.