[Gretl-users] distribution of equality-of-means
Today I noted that for an equality-of-means test assuming unequal
variances, gretl uses t(n1 + n2 - 2) as the relevant distribution. I'm
not sure right now, but isn't this only ok with equal variances? So that
the test is only justified asymptotically with standard normal
distribution (through a central limit theorem)?
Thanks again,
sven