Hi,
On Thursday, 2007-03-08 21:36:18 +0100, Eike Rathke wrote:
> The CHITEST function still lacks a definition, anyone?
While defining that, I stumbled over some odd Excel behavior. Excel
allows data arrays (actual and expected values) of different dimensions
as long as they have the same number of elements
(rows_actual*cols_actual==rows_expected*cols_expected). To me that
doesn't make sense and I wouldn't know how to fit that into some
Chi-square formula other than treating the entire data set as one and
not in different columns. It especially doesn't fit into the formula
that's given in Ecma/Excel. IMHO.
Furthermore, Ecma/Excel's definition of degrees of freedom looks weird,
for rows>1 and cols>1 they say it would be (rows-1)*(cols-1). Why?!?
Can anyone shed some light on this?
Eike
--
Automatic string conversions considered dangerous. They are the GOTO statements
of spreadsheets. --Robert Weir on the OpenDocument formula subcommittee's list.
