Se() funcitonality for Dirichlet regression // measurement error

Interfaces brms
1

Are there plans to implement the resp_se function for Dirichlet regression?

As in:

brm(y | se(x) ~ x,
    family = dirichlet(link = "logit", link_phi = "log"))`

I am trying to compute a time-series of variables that sum to 100 but also have (sometimes) significant measurement error. Maybe there’s another way to do this?

Related question: can I specify a number in the se() function to pass the model a predetermined standard error?

Thanks,
Elliott

  • Operating System: MacOS 10.14
  • brms Version: brms 2.10.0
2

See https://github.com/paul-buerkner/brms/issues/754 for my answer. Please avoid cross posting questions even if I may not reply for a few days. I will eventually reply to all brms questions asked here on discourse.