After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 68057 - F test significance gives #VALUE
F test significance gives #VALUE
Status: RESOLVED FIXED
Product: Gnumeric
Classification: Applications
Component: Analytics
git master
Other other
: Normal normal
: ---
Assigned To: Andreas J. Guelzow
Andreas J. Guelzow
Depends on:
Blocks:
 
 
Reported: 2002-01-04 23:14 UTC by Adrian Custer
Modified: 2004-12-22 21:47 UTC
See Also:
GNOME target: ---
GNOME version: ---



Description Adrian Custer 2002-01-04 23:14:04 UTC
In the regression analytics, I'm getting #VALUE immediately below
"Significance of F". If this is the p-value of the F stat, R gives a real
value for the same data.

Also, would it be possible to add a checkbox "expand columns to fit"? For
new sheets it would be cool if I didn't have to re-size columns right after
making the regression test. 

cheers,
adrian
Comment 1 Andreas J. Guelzow 2002-01-05 23:01:06 UTC
adding an expand column to fit checkbox is probably a good idea for
all analysis tools. Currently some autofit some don't.

About #Value: Does this happen to you all the time? Could you attach a
sample sheet that couses this problem? Which valuedoes R calculate for
that data?

Thanks
Comment 2 Andreas J. Guelzow 2002-01-05 23:02:29 UTC
If it were possible, I would also reassign this bug to me.
Comment 3 Adrian Custer 2002-01-06 17:10:22 UTC
1) new gnumeric
2) B4:b13 gets 1,2,3...10
3) C4:C13 gets 23,43,23,43,45,56,67,98,...
4) select a1
5) Tools->Data Analysis -> regression
6) xvar gets B's
7) yvar gets C's
8) ok
--> in new sheet regression the #value appears.

did jody change this needinfo->new because he got this also?


cheers,
adrian
Comment 4 Andreas J. Guelzow 2002-01-06 20:54:59 UTC
I can repeat that (even if I don't quite know what the y values really
are).

The bug really in the calculation of finv since even the finv function
has this problem:

=finv(0.025,1,8) evaluates to #VALUE!
Comment 5 Andreas J. Guelzow 2002-01-07 14:02:26 UTC
a diff implementing the enhancement of an `autofit columns' checkbox
in all analysis tools has been submitted to Jody.

finv(0.025, 1, 8) gives 7.570882 on Morten's Sparc system.
Comment 6 Morten Welinder 2002-01-07 14:46:47 UTC
Initial debugging suggests that this is caused by "gcc 2.96" (red hat
or mandrake) miscompiling function qbeta in mathfunc.c
Comment 7 Morten Welinder 2002-01-07 14:54:34 UTC
Confirmed.
Comment 8 Andreas J. Guelzow 2002-01-07 21:24:11 UTC
Just to set the record straight: this was not a gcc 2.96 problem it
tuned out that most linux systems were not recognized as IEEE systems
and as a consequence some slightly buggy code was used. THe
configuration has been corrected so that w enow also look for
ieee754,h header files as replacement to ieeefp.h.

Moreover the requested autofit enhancement has been committed to cvs.