I have a measurement error model that I’ve specified along the following lines:
fit <- brm(
formula = y ~ me(x, sd.x),
...
)
Note that no values of x
are actually missing as such; they’re just measured with a certain amount of error. (In the actual model, they correspond to frequency estimates for various things, and I have estimates for every row.)
However, while looking up how to specify such a model from other sources, I discovered that the brms documentation says me
is soft deprecated in favor of using mi
. I looked at the documentation here, but am not entirely sure how I could translate my model to the new syntax. Is it as simple as this?
fit <- brm(
formula = y ~ mi(x, sdy = sd.x),
...
)
Or this?
fit <- brm(
formula = y ~ x | mi(sdy = sd.x),
...
)
Or is the new syntax not supported for my use case?