negative binomial regression

[clarkliming]

@mbannick I saw this link https://stats.stackexchange.com/questions/221648/negative-binomial-regression-in-r-allowing-for-correlation-between-dispersion

wonders if this is related?

[mbannick]

@clarkliming thanks for linking. I don't think that this impacts what we should do for RobinCar2. This is related to the variance-covariance matrix of the regression coefficients. We wouldn't use that when constructing the variance-covariance matrix on the outcome scale using the vcovG. The sandwich vcov would, but we don't want to implement that for negative binomial regression, only for linear models in the special case.