Thanks.  I pushed a commit that allows +/- .001 for this result.

John Darrington <[email protected]> writes:

> I've noticed this on GNU/Linux too, depending on what optimisation 
> flags are used.
>
> I don't think one digit in the third decimal place is a big deal.
> Especially as the p-value is unchanged.
>
> On Sun, Jul 10, 2011 at 03:57:19PM -0700, Ben Pfaff wrote:
>      
>      406 fails due to a numeric precision issue.  John or Jason, would
>      you mind taking a look to verify that it's not a big deal?
>      Here's the report:
>      
>          406. glm.at:3: testing GLM latin square design ...
>          ./glm.at:59: pspp -O format=csv latin.sps
>          --- -   2011-07-06 18:28:20.000000000 -0400
>          +++ 
> /opt/local/var/macports/build/_Users_aeetes_dports_math_pspp-devel/pspp-devel/work/pspp-0.7.8-g42ec6a/tests/testsuite.dir/at-groups/406/stdout
>       2011-07-06 18:28:20.000000000 -0400
>          @@ -7,5 +7,5 @@
>           Factor C,155.596,5,31.119,9.350,.000
>           Error,66.563,20,3.328,,
>           Total,1144.730,36,,,
>          -Corrected Total,329.627,35,,,
>          +Corrected Total,329.628,35,,,
>      
>          406. glm.at:3: 406. GLM latin square design (glm.at:3): FAILED 
> (glm.at:59)
>      
> J'

-- 
Ben Pfaff 
http://benpfaff.org

_______________________________________________
pspp-dev mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/pspp-dev

Reply via email to