Morten,
Oh, kudos to the folks who implemented stdev, var, binomdist, and similar functions: they go far in fixing things within Excel that are still buggy, after all these years. (sheesh, sounds like a song...) That wasn't really hard. In fact, at my university we used to teach first-year students not to make those mistakes.
I *know* it isn't hard; I just wish the powers-that-be responsible for the stat functions had been among your or my students <g>. BTW, it is my (informed) understanding that a lot of that was done purposely by MS in the early days to replicate what Lotus 1-2-3 produced; that the mistake was intentionally coded in. I'm hoping that those that can move Gnumeric in the right direction on the stuff I posted earlier don't adopt the same attitude, namely "the most widely-used package is messing up, so let's follow suit".
We may not have gotten everything right, though. Do speak up if you find something stupid like negative variances.
I will definitely do so. As I get more familiar with your project and that of R, I will be happy to help out with the algorithms as well.
Regards, Dave