[SciPy-user] array mean issue...

fred fredmfp@gmail....
Wed Jun 11 07:08:20 CDT 2008

Sebastian Stephan Berg a écrit :
> Hello,
> This is an overflow problem in the way the mean is computed I believe --
> it would start to add nothing after a while. In any case you can get the
> correct/better result with b.mean(dtype='float64').
Thanks, Sebastian.

However, I use nansum to compute the mean.
nansum/size gives the same wrong result as mean(), and it seems that is 
not possible to set the dtype with nansum.

Any clue ?




More information about the SciPy-user mailing list