I've got a continuous y (bounded between 0 and 10) and a 3 level categorical x (small, medium, large) and some other covariates (age, gender).
After OLS regression, both the residual vs fitted plot and Breusch-Pagan / Cook-Weisberg test for heteroskedasticity seem to suggested heteroskedasticity isn't a problem.
However if you look at the rvfplot, it looks like this (red line of best fit, green lowess).
Firstly, am I correct to surmise that heteroskedasticity is not an issue with this model, even though the plot looks odd. Secondly, how does one explain the rvfplot?