Different Output to Stata

[hanjo-kim]

Hi, not sure if this warrants an issue, but I think this may need some clarification.

I successfully replicated your example in Julia, but I wanted to make sure it matches up with Stata output as well.

In Stata, after reading the dataset, I do the following:

xtset state year
xtreg sales ndi i.year, vce(cluster state) fe

The beta coefficient turns out to be exact compared to Julia's output, but the standard error in Stata is 0.0025274 whereas in Julia, it's 0.0025283, which lead to a different t-statistic and confidence interval.

F statistic also seem to be different (Julia 7.327 vs Stata 62.26), as well as R-sq within (Julia 0.154 vs Stata 0.4594)

I am using the same dataset, so I would think that they would be the same. Is there any reason as to why this might be? I think it would be incredibly helpful to match the result of Stata exactly to make sure that there are no unforeseen errors that might pop up in the future? Or perhaps, I am missing something about the procedure.

Anyway, it would be highly appreciated if someone told me what I am perhaps doing wrong?

Thank you in advance.

[hanjo-kim]

So I used reghdfe, and now the results are much closer. There seems to be a difference the degrees of freedom or effective observations that makes standard errors be slightly different. There must be something going on in the background that I do not understand. I will close this issue.