Re: [office-comment] Statistical SUMMARY() function

["David A. Wheeler" <dwheeler@dwheeler.com>]

Leonard Mada:
> One of the most used features in spreadsheets is to analyze and 
> summarize the data. This is used in academia and businesses alike.
> 
> In both instances, spreadsheet users often need to perform an 
> exploratory data analysis and in one of the first steps they will 
> summarize their data.
> 
> I discussed this issue more fully at:
> http://www.openoffice.org/issues/show_bug.cgi?id=89980
> 
> While Calc (and ODF) has (most) functions implemented as separate 
> funcitons, it lacks a unified function to compute all parameters at 
> once. MS Excel does come with such a 'Menu'-entry in the advanced 
> statistical toolpack.

I agree that statistical summary analysis of a dataset is useful.

I'm not convinced that this is best done by a function, though.
There's no way to know how many rows it will create, for example,
which will cause many problems for most implementations.

This looks a lot more like the graphing capabilities than a function.
E.G., you can set the range of values, and then auto-generate
a separate summary report.  Instead of a graph, it'd be a
display of the statistical information.

I agree that if reports aren't auto-calculated
you can end up with misleading results.  But that is true for
graphs too.  I think it'd be better to think about this being similar
to the graphing capabilities, and going from there.

I think it's hopeless to enter _this_ round anyway.  There are no
implementations, and that's usually a killer unless it's easy
to implement, clear, and solves a problem that's widely agreed to.
Other than IFERROR() and NUMBERVALUE() I don't recall adding
anything to the spec that didn't have at least one implementation,
and those two had simple semantics and clear value.

--- David A. Wheeler