MultiComparison w/ Pandas Series
Javier Burroni
In [113]: np.unique(groups.values, return_inverse=True)
Out[113]: A
(array([u'A', u'B', u'C'], dtype=object),
array([2, 0, 0, ..., 2, 2, 1]))
In [114]: np.unique(role, return_inverse=True)
Out[114]:
(role
3641 NaN
59108 NaN
43653 NaN
1581 NaN
75610 C
55740 NaN
16712 A
75614 NaN
102468 NaN
55742 NaN
102470 NaN
30488 NaN
102472 C
30489 NaN
4962 NaN
...
38778 NaN
38779 A
38781 NaN
38782 C
38784 NaN
38785 NaN
38786 A
38787 NaN
38789 NaN
38790 NaN
38791 NaN
38792 NaN
38793 NaN
38794 B
38795 NaN
Name: role, Length: 105590,
array([68079, 21442, 5820, ..., 77651, 77652, 24166]))
" To be is to do " ( Socrates )
" To be or not to be " ( Shakespeare )
" To do is to be " ( Sartre )
" Do be do be do " ( Sinatra )
josef...@gmail.com
> hi,
> I wanted to perform MultiComparison using pandas series but it fails.
> mod = MultiComparison(x, groups)
> this statement never ends (or it doesn't in few minutes).
> But, if I do this
> mod = MultiComparison(x.values, group.values)
> it works perfectly.
>
> Apparently the issue is with unique, inv = np.unique(groups,
> return_inverse=True) which gives a different value for unique as using
> unique = np.unique(groups).
>
> I have no created an issue as I couldn't make a synthetic example. Also, I
> don't know to whom report. Most probably this is not the place and it should
> be reported to pandas, but as this affects pystatsmodels behavior, I'm
> writing this email
your message in the description.
The class doesn't convert the data to arrays, that's a bug in MultiComparison.
We might want to use more information that we can get from a pandas
data structure, but this was written before pandas or without pandas
in mind. (If groups are a pandas factor, then we don't need to find
the levels again.)
related for weightstats https://github.com/statsmodels/statsmodels/issues/771
(A warning: except for tukeyhsd and it's supporting code, there is not
much test coverage for that module. 23% only)
Thanks for reporting
Josef