What is the difference between using poisson or glm with family set to poisson?

[kyle dack]

I have a poisson regression model, and I understand Stata can fit it using:
1. poisson cases var1 var2 var3, exp(expected_cases)
or
2. glm cases var1 var2 var3, exp(expected_cases) f(poisson)
Both produce the same coefficients and other results. But I noticed I can only diagnose residuals using rvfplot2 after the second command, not the first. This made me curious whether there is any other difference in using the two commands, and whether one is preferable?

[Richard Williams]

glm can estimate many models that more specialized commands can also estimate. When that is the case, I think the specialized command is usually preferred. The code is better optimized, and the options and post-estimation commands will generally be better tailored to the command. But, as in the example you cited, there may be a few occasions where glm has advantages.