[Perldl] Should NaN be missing value or error value?

chm devel.chm.01 at gmail.com
Sun Aug 14 11:12:37 HST 2011


Ironically, the "old" pre-badvalue code already
implements the planned behavior.  Some additional
testing needs to be done with and without badvalue
support in PDL.

--Chris

On 8/14/2011 3:54 PM, chm wrote:
> PDL Users-
>
> The current implementation of maximum, minimum,
> and the related operations are inconsistent
> with respect to NaN values in the input data.
>
> As long as the first element is a number, then
> the routines treat NaNs as missing values. If
> the first value is NaN then the output is NaN
> which is inconsistent with the former since the
> result will not be NaN unless all the values are
> NaN.
>
> I would like to fix this so the result is
> consistent and am leaning towards the missing
> value option. That is (1/dim(0)) of what
> we were usually calculating. :-)
>
> Please reply with your concurrence or any
> other thoughts/viewpoints.
>
> Thanks,
> Chris




More information about the Perldl mailing list