Scale parameter in cumulative logit models


#1

Hi,

I am fitting a cumulative logit model to survey responses about illegal fishing activity.

My model call is:

M.SR <- brm(formula = Score ~ Species + Region + (1|ID), data = IUULong, family = cumulative)

Where Score is a ranking of the level of illegality from 1 to 5, Species and Region are self explanatory. ID is the respondent id in the survey.

The respondents have ranked their experience with particular species as part of the survey. I would like to incorporate the experience ratings into the model, likely as part of the scale for the latent response variable.

For an example, see section 4 in the vignette for CLM in package ordinal. Its at https://cran.r-project.org/web/packages/ordinal/vignettes/clm_tutorial.pdf

I would happily provide some R code, but I have not been able to figure out how to get a scale parameter into the cumulative family coding.

I am working on a mac with brms 2.2.0

Thanks

Chris


#2

Try out

bform <- bf(
  Score ~ Species + Region + (1|ID),
  disc ~ <your predictors>
)
M.SR <- brm(bform, data = IUULong, family = cumulative)

where disc is short for discrimination happens to be the inverse of the scale parameter you want to model. :-)