Can I use a Negative Binomial Regression for a Continuous Outcome?


My outcome is Sugar intake with a highly skewed continuous outcome variable. The histogram and variable details are Attached.

Since the distribution is highly skewed and has a very high variance (12.78) compared to the mean (2.67) I thought I could round each decimal to the nearest integer, treat this as count data and use the Negative Binomial Distribution. However, I am now second guessing myself. I have looked at multiple transformations, but none seem to fit.

My main question is “Is the Negative Binomial Distribution the best distribution to use for this Outcome? And if not, do you have suggestions on which distribution I can try?”

I did read about a gamma distribution and used the gammafit command in Stata, but am not sure how to interpret the output or if I should even use it.

Thank you
Last edited: