Fitting silently fails for a series of type object

[Dov Grobgeld]

When trying to do linear regression of a pandas.Series() object I used the following code:

import pandas as pd
import statsmodels.formula.api as sm

x = [0.,1.,2.,3.,4.]
y = [1,2.1,2.9,3.1,3.9]
s = pd.Series(y,index=x)
df = pd.DataFrame({'Y':s,'X':s.index})
fitted_values = sm.ols(formula='Y ~ X', data=df).fit().fittedvalues

I was suprised that the fitted_values are identical to y:

In [14]: fitted_values
Out[14]:
0 1.0
1 2.1
2 2.9
3 3.1
4 3.9

This seems to be due to the fact that s.index has the dtype 'object'. If casted to float64, this solves the problem:

df = pd.DataFrame({'Y':s,'X':s.index}).astype(float64)
fitted_values = sm.ols(formula='Y ~ X', data=df).fit().fittedvalues

In [19]: fitted_values
Out[19]:
0 1.24
1 1.92
2 2.60
3 3.28
4 3.96

In my opinion, either sm.ols should do this type conversion internally, or throw an exception about illegal input.

Version of statsmodels is 0.5.0 .