I have a fractional dependent variable and I am using a fractional heteroskedastic probit model. However I am not not sure as to how to decide which variables to include in the variance equation.
(I am using a cross-sectional household survey and I have included state-level variables also in the regression. I have three state level variables and when I include two of these state level variables in the fractional probit regression my main results reverses in sign-which made me consider a fractional hetprob model). Also there seems to be no multicollinearity as the vif value is low.