[SciPy-dev] Stats t-test broken on 1D input

Ed Schofield schofield at ftw.at
Thu Mar 2 16:13:47 CST 2006


On 02/03/2006, at 8:50 PM, Zachary Pincus wrote:

> Hi folks,
>
> Sorry to re-post, but nobody in scipy-user seemed too concerned that
> the most basic form of the most basic statistical test -- the T-test
> on 1D input -- is broken in scipy 0.4.6.

Thanks for your insistence in posting this.  I've now fixed it (I  
hope) in SVN.  I actually think this is an old bug.  We don't have  
ANY unit tests for the T-test (or normality test, etc.), and I think  
you've stumbled across it because you're the first one to have used  
it since long, long ago ...

The ttest_ind function actually has a comment that it's "from  
Numerical Recipes".  So I think, even though it's been Pythonized, we  
don't have permission to distribute it and we need to need to remove  
it altogether.  So we really need some more help with the stats  
module ...

-- Ed




More information about the Scipy-dev mailing list