Hi, I’ve been running a hierarchical multinomial logit model using the Stan code shown below. I’m finding that the model tends to be overfitted when comparing in-sample and holdout accuracies (for each respondent I hold out responses to some questions and fit the model on the remaining questions, t…

I’ve done some previous work with multilevel, multinomial models in Stan. It would be helpful to learn more about the research and data structure. It sounds like you have data on respondents who make multiple choices throughout the study. Can you tell us more?

You could add a random intercept for each observation. Second use a noise nugget to your covariance matrix. Your covariance matrix \Sigma would become \Sigma + \sigma I, with I identity matrix. No diagonal update in the cholesky form is known, thus sampling the full matrix and doing a cholesky d…

Yes, the data is from a conjoint analysis where a number of survey respondents are asked to respond to a number of questions in which they choose between a number of alternatives, each with different attributes. The matrix beta contains the parameters (utilities) for each respondent * attribute, X c…

You might benefit from this thread (though it doesn’t directly relate to your question about tuning parameters): [image] Mixed Logit Model Modeling Hi Before setting up my own mixed logit model I wanted to ask if such a model already exists. By mixed logit I understand a …

Thanks Andre, I’ll look into your suggestions!

Overfitting in hierarchical multinomial logit model

Modeling

jeremy.koster November 8, 2018, 5:50am 5

You might benefit from this thread (though it doesn’t directly relate to your question about tuning parameters):

Topic		Replies	Views
Hierarchical multinomial model Modeling	8	4280	February 8, 2018
Slow Stan Code for Hierarchical Multivariate Model Modeling	8	1377	July 24, 2018
Speeding up a hierarchical multinomial logit model Modeling	27	8312	October 18, 2022
Hierarchical Multivariate Normal model Modeling fitting-issues , performance	6	2255	June 12, 2019
Hierarchical overdispersed count models Modeling binomial-response , count-data	2	868	November 10, 2021

Overfitting in hierarchical multinomial logit model

Related topics