Somersd function from scipy.stats produces type error in Python

I would like to calculate the somers d. I have the following input cross table in python as a data frame called crosstab:

0	1	2	3	4
100	80	100	4	500
50	3	2	0	38
40	0	4	0	40
2000	100	100	4	400

I try to calculate the sommers d and get the following error message:

from scipy.stats import somersd
somersd(crosstab)

TypeError: Invalid call to pythranized function `_concordant_pairs(int64[:, :] (with unsupported column-major layout))' Candidates are: - _concordant_pairs(int[:,:]) - _concordant_pairs(float[:,:])

Where is my mistake?

Solution

You need to use to_numpy because x in somersd should be an array-like of floats :

out = somersd(x=crosstab.astype(float).to_numpy())

Output :

print(out)

SomersDResult(
        statistic=-0.5710103041724046,
        pvalue=9.019579189415401e-226,
        table=array([[ 100.,   80.,  100.,    4.,  500.],
       [  50.,    3.,    2.,    0.,   38.],
       [  40.,    0.,    4.,    0.,   40.],
       [2000.,  100.,  100.,    4.,  400.]]))

Input used :

crosstab = pd.DataFrame(
    {0: [100, 50, 40, 2000],
     1: [80, 3, 0, 100],
     2: [100, 2, 4, 100],
     3: [4, 0, 0, 4],
     4: [500, 38, 40, 400]}
)