[SciPy-dev] Deprecating current scipy.stats.var API for 0.7 ?

[David Cournapeau]

Anne Archibald wrote:
> Consistency sounds like a good idea to me - though having the same
> functions under different names is not much more consistent than
> different functions under the same name.

I disagree on this particular case at least: the axis difference is 
annoying. I would not mind removing the function at all if we can 
provide all the wanted behavior in numpy.

>  But if the functions in
> scipy.stats are the same as the functions in numpy, why have them at
> all? It seems like they should present some increase in functionality.
> We have had a number of requests for versions of mean, std, and var
> that are more numerically stable; perhaps these should be the new
> ones?
>   

Is there any reason why the numerically stable cannot be put into numpy 
proper ?

cheers,

David