Implementing a truncated regression for a normal distribution in R

I'm not a statistician but I'm working with some experimental psychology data. I have a distribution of responses on a -4 to 4 scale. Usually, these type of variables is treated as continuous. I have to model this variable based on some predictors (experimental factors). I'm thinking about different ways to model it:

• Linear Mixed-Effect Model (for repeated measure)
• Ordinal Regression (if I consider the variable as an ordered factor)

However, I've heard about the possibility to use a Truncated Normale as target distribution. Could be useful in this case? Is there a way to do it in R?

1 Answer

You could also use a Beta mixed effects model. In this model the response data are in the $$(0, 1)$$ interval. To fulfill this requirement, you could transform your original data $$Y$$ into $$Y^* = (Y + 4) / 8$$ and fit the Beta mixed model for $$Y^*$$.

You could fit this model in R using the GLMMadaptive package. For examples, look here.