[Numpy-discussion] Question about masked arrays

[Benjamin Root]

On Mon, Sep 20, 2010 at 4:03 PM, Gökhan Sever <gokhansever at gmail.com> wrote:

> On Mon, Sep 20, 2010 at 3:34 PM, Benjamin Root <ben.root at ou.edu> wrote:
>
>> I have been using masked arrays quite extensively.  My take on them is
>> that if a masked array makes sense in that operation, then they should still
>> work with the regular functions.  However, there have been many cases where
>> a developer used np.asarray() instead of np.asanyarray() for their code,
>> which causes the masked array object to lose the mask.  If you encounter
>> such situations, it is usually a bug and should be reported.
>>
>
> Do you think then in my case assigning the masked array to np.zeros and
> getting a non-masked return is a bug or I should make sure that I am
> providing a masked array equivalent zeros function before I start
> doing computation?
>
>
If you have any variable and you assign np.zeros() to it, then whatever was
there previously is replaced.  If you want to create a masked array, then
use np.ma.zeros(), or take an array and put it through np.ma.masked_array()
with the mask keyword to create the mask at the same time.  Note that this
approach supposedly uses the original array, so you aren't paying much of a
penalty here.


> And also if the inner execution could be clarified by asanyarray assertion
> why there is ma equivalent array operation functions?
>
>

That is a design question for the numpy gods...


>
>> Actually, that reminds me... watch out for np.polyfit() with masked
>> arrays.  It doesn't behave quite nicely with masked arrays and the results
>> are deceptive.  It may appear to be right, but it is not.  Use
>> np.ma.polyfit().
>>
>
> Hah, you should see my scipy.optimize.leastsq in action. I estimate
> geometric mean and standard deviation parameters for log-normal aerosol
> population. It runs, with the values masked and unmasked (from above
> np.zeros and np.ma.zeros difference) but resulting all kind interesting
> estimates.
>
>

I wonder if the functions were originally intended to work with masked
arrays, but at some point, some inner part of the algorithm became
'ma-unfriendly'...

Ben Root
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20100920/fcce8176/attachment.html>