Re: Calc statistical accuracy

> This is a question more from curiosity than anything.
>
> Today I was reading that the Journal of Computational Statistics and Data
> Analysis was not at all happy with Microsoft Excel 2003, stating that:
>
> "Given Microsoft’s track record, we recommend that no statistical
> procedure be used unless Microsoft demonstrates that the procedure in
> question has been correctly programmed, e.g., by the use of test datasets
> from textbooks, or by comparison with another software package." [1]

I would agree with that for all spreadsheet programs. There is statistical
software out there that has been tested and is known to produce good
results for that like R.


> What interests me the most is that they gave some specific advise on how
> to test statistical software. In particular they stated that NIST has a
> Statistical Reference Dataset that can be used to test statistical
> software. [2]
>
> I found the datasets on the NIST website [3]. They have a number of tests
> for ANOVA, Linear Regressions, Markov Chain Monte Carlo, Nonlinear
> Regression and Univariate Summary Statistics.
>
> I was wondering if we've done any work on testing Calc against these
> datasets?
No and as long as you don't plan to work on them I doubt anyone is going to
do it soon.

Additionally as long as this mainly serves as theoretical discussion I
think it is off-topic here and should be moved to the discuss list. However
if you want to work on improving Calc's statistical capabilities I think
the best start is to think about a interface to connect it to R. I think we
even have somewhere at some old wiki page this idea as a crazy idea for the
future.

Regards,
Markus