Why do we say that with conditional heteroskedacity, the standard error is underestimated?
Technically, i think standard errors can be overestimated, and our T and F values will result smaller. As a consequence, we wouldn’t reject the H0 when we would should have and that would cause a type II error
Why do you think that the standard errors would be overestimated (i.e., too big), rather than underestimated (i.e., too small)?
Just a hunch, or do you have some analysis to back up that thinking?
it was just a hunch
i cant seem to understand why the standard errors will be too small if we have more variations with indipedent variable changing
anyone?