What does BRMS do when missing values in only one variable of multivariate model (mvbind)?

I am running a multivariate regression in BRMS, with 2 variables. Some participants have missing values for one variable, but not for the other. If I run the analysis using mvbind as in https://cran.r-project.org/web/packages/brms/vignettes/brms_multivariate.html, does the model just completely exclude all data for a participant with one missing value? Or will it use their data for the one variable where they have an observation, and exclude them for the model where they do not have an observation?

Relatedly, I saw a discussion about possibly adding in the ability to impute missing values and incorporate this into a model. Is this something that has now been implemented?

Any help would be much appreciated!

Please also provide the following information in addition to your question:

  • Operating System: Mac OS 10.12.6
  • brms Version: 2.8.0

There is now a brms vignette available about missing values: https://cran.r-project.org/web/packages/brms/vignettes/brms_missings.html


Excellent - thanks for your quick response!

@paul.buerkner the vignette is very helpful, and I will probably now go and try the multiple imputation myself. Just so that I can properly understand the output I already have from using mvbind without the approaches used in the vignette, is there a simple answer to whether or not the model is fully ignoring people with one variable missing, or whether they will contribute to the model but only for the variable where they have an observation?

By default brms will do rowwise exclusion if one or more variables are missing in a row.

Thanks again - I’ll take a look at doing some imputation!