I have a question regarding how to deal with missing values in zero-inflated models. In these kind of models, we have discrete values (zeros) which come from Bernoulli distribution and non zero values which come from either continuous or discrete distribution.
I saw the concept of marginalization in some forums. But I don’t understand how can I apply it for this kind of models. I’m grateful if anyone have any solution for this.
Thank you very much
can’t you do multiple imputations where you also impute the missingness in the outcome?
Thank you for responding soon. I’m just wondering how can I implement in Stan. Because my missing values contains zero as well as non-zeros. But I cannot specify it as a parameter in the parameter block since it contains discrete values. I would like to know if there’s way that I can implement it in Stan.
sorry, only done it with
brms and the
mice package. Here’s a good introduction:
But, if you have zeros and NAs you first need to decide if the zeros are actual zeros, and not NAs coded as zeros.
Thank you very much. I appreciate your feedback. I’m interested in coding it using Stan.