I have an independent variable grouped into irregularly sized classes (0, 1-5, 6-10, 11-20, >20), a continuous dependent variable, and several control variables.
So far I've been trying to stay close to the format of the observed data, treat the IV as categorical and run a One-way ANCOVA. Variances were unequal, so I opted for a mixed linear model instead. This however feels overly complicated for such a simple problem and I was considering to instead just transform the data into class midpoints (i.e., 0, 3, 8, 15.5, ?20?) and treat the IV as continuous to perform a simple regression instead. Would this be a valid approach? Are there any alternatives that might be more suitable?
Thanks in adcance