Good question. One approach is to fit an independence model repeating the same individuals over time and making sure each is represented in the same cluster ID in the survey design. This approach would estimate a “robust” sandwich covariance matrix. It treats the repeated measures as a nuisance correlation, but the regression coefficients and other parameters have the variance adjusted. If on the other hand, we’d want to do a variance decomposition and treat the individual as a group level random effect, this gets trickier and we’d probably want to weight the random effect distribution for the individual, leading to multiple weights. I don’t think it can be directly done in brms and adjusted. We tried the estimation part in a couple papers and have recently tried the adjustment for a simple one-way ANOVA. The cs_sampling adjustment didn’t work “out of the box” and had to be tweaked. If you have a motivating example, we can try to work through it!