Hi,
As you may now, both smf.ols and mf.glm can be used for fitting a regression model. Theoretically, they should bear the same results if the error is normally distributed. I ran 2 regressions (attached please find the code), once using ols and the other using glm. While, their results are identical, ols uses t-test but glm uses z-test. Consequently, their confidence intervals are different.
To my knowledge, we always should use t-test since we do not know the actual value for sigma and we estimate that by the samples we have from the population. I was wondering if someone can explain why smf.glm uses z-test.