This pertains more generally to my growing amateur interest in numerical analysis and approximation rather than HP calculators per se. Since such matters are of huge importance to some who program HP calculators I hope the group will indulge me one more dalliance.

In my efforts to learn more about the practical aspects of minimax approximation I have been experimenting with the Solver applet in Excel. Now I know it is not the most sophisticated optimization tool out there, but it is fair enough for my low level messing about at this juncture.

What I have discovered in passing is that whereas some of Excel's built-in functions (like the trig, log, and exponential operations) have no problem rendering full 15-digit precision if that is what you want, some other special functions, like ERF(), ERFC(), and certain of the statistical distributions I have experimented with, don't seem to offer such high precision--pretty good for some arguments but clearly approximations and not to 15 digits. Therefore, if I want to generate approximations based on these functions I need to import full precision data points from some other software. Fair enough. If that is what is necessary there are ways I can do that.

But my question is this: Does Microsoft or anyone else publish for the more mathematically sophisticated end-user any specifications or data on such functions, the algorithms used to approximate them, and estimated error, be it relative or absolute? I must admit that I really put trust in the precision of the Excel functions until I had trouble replicating fits done elsewhere, and then noticed that my supposed full precision comparator was not so precise at all.

I hope this query generates thoughtful replies. I am learning that serious HP users seem to have a broader interest in software, programming, and numerical analysis, so this admittedly general question shouldn't seem to out of place here, I hope!

grateful and eager, as always,

Les