Co-variance matrix is undertermined

Hi all,

I am writing to you since the following error appears when using:

est = LinearDMLCateEstimator(model_y = LassoCV(cv=[(fold00, fold11), (fold11, fold00)]), model_t = MultiTaskLassoCV(cv=[(fold00, fold11), (fold11, fold00)]), n_splits = [(fold0, fold1), (fold1, fold0)], linear_first_stages=False)

Then I fit in this way:

est.fit(Y, T, X[:, :(n_products)], W, inference='statsmodels')

The error is:
/Users/federiconutarelli/anaconda3/envs/pharma/EconML/econml/utilities.py:961: UserWarning: Co-variance matrix is undertermined. Inference will be invalid!

Should I specifiy the covariance matrix somehow? What does it mean undetermined? I mean, in a mathematical sense or in the sense that I should set a specific covariance matrix?

I am operating in a panel data framework. Let's say I am almost reproducing the code of issue #94.

Thank you,

Federico

Start with the warning at econml/utilities.py:961 and reproduce the LinearDMLCateEstimator example using statsmodels inference. Determine whether the covariance warning is expected for this panel-data setup or indicates incorrect handling, then document the result or define the needed correction.