I’m trying to fit a EIV model but with one caveat – I know the measurement error variances (of all variables) but want to work on the log scale. So if the ordinary formula would be

`F_obs | mi(F_sd) ~ me(M_obs, M_sd)`

I would be interested “in sth like”

`log(F_obs | mi(F_sd)) ~ log(me(M_obs, M_sd))`

Would it be possible to fit such model with brms or should one instead treat the data as from a log-normal distribution and use the relation between log-normal distribution and normal distribution to form it as a model without logs.