Thanks Jerry! Btw, your function is a LOT faster than Excel's tinv!

But the results are still the same.

I'm beginning to wonder.. maybe the distribution of the sample kurtosis
is so *terribly* skewed, even for sample of size 60000, that it looks
to me like I'm getting a biased result? And maybe the variance is so
terribly huge too (I only know the theoretical variance of the sample
kurtosis for normal, not t distribution)..