Allow sm.OLS to work with Float64 values

[marcdelabarrera]

When working with Float64 values form pandas, sm.OLS raises an error

ValueError: Pandas data cast to numpy dtype of object. Check input data with np.asarray(data).

Example

impor pandas as pd
import statsmodels.api as sm
x = pd.Series([1,2,3,4]).astype('Float64')
y = pd.Series([1,2,3,4]).astype('Float64')
sm.OLS(y,x).fit()

It is easy to convert the Float64 into a float64 value, but it would be nice if sm.OLS can handle that.

[bashtage]

There is no support for extension arrays. While in your case it is simple to convert, how should pd.NA and np.nan be handled in an array that contained these?