I am interested in fitting a latent variable model to a set of data. I have played with the priors quite a bit and made them very tight, and have adjusted the delta and treedepth, and increased iterations to a high number, but I am not having any luck with convergence. If anyone can help me take a l…

I am afraid the model will remain unidentifiable. Without any constraints, different values of loadings and scores will give the same linear predictor and thus the same log posterior. This will lead to multiple modes in the posterior and impede you to compute posterior mean and uncertainty for the s…

Have you already looked at resources like this? [image] Confirmatory Factor Analysis using brms brms Hi all, Not a problem, more a little bit of helpful code. brms can already estimate latent variables using the mi() function for missing data. It can also constrain param…

Thanks for the response! The Scott Claessens post is actually what I used as a template for specifying the model. But I just wasn’t sure if I had correctly incorporated the random effect in the right place, because the Claessens example does not have one.

I’m far from an expert here, but in both cases they constrain parameters with either the constant() or a “stupidly narrow” prior. Yours seem much broader. Again, not sure what the solution is in your case but something to consider.

Yes, I noticed that some of the coefficients were set to be constant at 1 in the examples, but I did not think that constraining any of the parameters to be constant was appropriate in this case because I am not trying to do confirmatory factor analysis. I am actually trying to use the data to estim…

Thanks Lucas! Just to confirm because I am not familiar with the terminology of factor analysis, the term “loading” here refers to the slope relating the latent variable to one of the observed values, correct?

Yes, the model is at the end a CFA, that is what the Scott Claessens is replicating. The 2 most common constraints would be to set a factor loadings (regression between latent variable and item), or the variance of the latent variable to 1. These do not change the model, but change the point of refe…

I wanted to follow up and say that I did eventually achieve convergence but only when I set residual correlation to FALSE and constrained two of the parameters to be constant at 1: sigma_TWtrue, the variance of the unobserved true value of total weight, and b_Vol_miTWtrue, the slope relating the uno…

Also, thanks again for your help, I wish I could mark both as solutions!

Issues with convergence for latent variable model in brms

Modeling

franzsf June 1, 2022, 2:51pm 2

Have you already looked at resources like this?

Topic		Replies	Views
Confirmatory Factor Analysis using brms brms	9	2607	May 4, 2025
Brms model does not converge - any other solutions than stronger priors? brms rstan	7	1950	December 3, 2019
Convergence failure (maybe) in brms brms	8	2790	May 10, 2018
Multivariate latent variable model (factor analysis) in brms brms brms	3	1176	November 28, 2023
Convergence issues with multivariate model in brms General fitting-issues , brms	2	102	January 20, 2026

Issues with convergence for latent variable model in brms

Related topics