ARMA models are not implemented for family 'categorical'

Interfaces brms
, ,
1

I’m trying to fit a multivariate repeated measures-mixed effects model in brms with one continuous predictor variable and one categorical predictor variable. It appears that it is currently impossible to fit this in brms? Below is my code:

mod<-brm(bf(Y1~X+(1|uniqueID))+gaussian()+cor_ar(~1|uniqueID)+bf(Y2~X+(1|uniqueID))+categorical(),data=mydata)

And I get the following error.

ARMA models are not implemented for family ‘categorical’

Is there any way to model a mixture of continuous and categorical dependent variables? I get a similar error message when I try with “bernoulli” and “Beta” families as well.

Thank you very much!

2

With your above model, this should indeed not happen. Please install the latest brms version from CRAN (2.12) ich you haven’t already done so. There, you can specify the autocorrelation structure direclty within the formula. For more details, see ?arma after installing the latest brms.

1 Like
3

Thank you very much Paul. This solves the problem.