CHIDIST parameter bug?

  • Thread starter Thread starter Cochran
  • Start date Start date
C

Cochran

I'm using Excel 2003 (SP2) in a spreadsheet reporting p-values
from statistical Chi-Squared tests. The sheet uses the CHIDIST
built-in statistical function to calculate the p-values.

The function CHIDIST( x, 999) will return a valid result only
if x is <= 971.18 or >= 999.00. But it returns a #NUM error if
x lies between 971.19 and 998.99. This occurs when the p-values
are between .4931 and .7301, which is right in the middle of the probability
distribution. This seems to be a bug with the product...there are no
limitations
on the parameters of the CHIDIST function described in the Help files, nor do
I find anything about this on the web. There is a description of fixes made
to other statistical Excel functions for Excel 2003, but this bug does not
seem to have been fixed at that time.

Any suggestions (other than to use a different product)?
 
Limited working range and limited accuracy of Excel's probability
distribution functions has been a topic of discussion for over a decade. In
Excel, the best solution is to download Ian Smith's VBA library of functions,
http://members.aol.com/iandjmsmith/examples.xls
which are world class for both accuracy and working range.

Alternately, you could consider doing probability calculations in either
Gnumeric
http://www.gnome.org/projects/gnumeric/
or spread32,
http://www.byedesign.freeserve.co.uk/
both of which are Excel compatible.

Jerry
 
Back
Top