Is it possible to account for censoring in predictor variables in brms? I have water pollution concentration data subject to left censoring (and measurement error, while we’re at it) that I would like to use as an exposure for a binary disease outcome.
I’ve considered whether censored predictors might be handled as a multivariate model, analogous to missing data in predictors, wherein:
bform_mi <- bf(y ~ mi(x1) + x2) + bf(x1 | mi() ~ x2) + set_rescor(FALSE)
bform_cens <- bf(y ~ x1 + x2) + bf(x1 | cens(x_cens) ~ x2) + set_rescor(FALSE)
but there is not the analogue of mi(x1) in the main model formula so I’m not sure what brms actually makes of that construction, nor if it’s appropriate to have the x2 as a predictor of x1 for the censoring portion.
Any guidance would be greatly appreciated!
Please also provide the following information in addition to your question:
- Operating System: macOS 10.14.1
- brms Version: 2.5.0