Hello everyone,
I am wondering why posterior only depends on the prior I gave and it seems the data does not work at all? My script can run pretty well without any warning issues…Thanks!
Can you provide some additional details about the model you’re running and the data you have? It would be ideal if you shared your model (or some version of it) so that others on the forum can verify that there isn’t something incorrectly specified within Stan. For example, it is possible that the likelihood statement for the data is not included, which would mean that the model is only fitting the priors.
Other possibilities could include the fact that you have some combination of overly informative priors and too little observed data. Depending on what you are using as priors and how much data you have, the posterior can look very much like the priors. Again, having the model code to look at would help to know whether this is happening or not.
You may try a posterior predictive check as well to see whether the posterior is actually predicting the data in a meaningful way. It could be that the initial information from the priors is just very close to what the true posterior is (which would suggest to me some highly informative priors).
Without additional information about your specific data and model, all we can do is speculate as to what might be the problem
2 Likes