Environmental correlation uses OLS, not Tobit

[jzuccollo]

Correlations with non-discretionary variables should, ideally, be estimated with a Tobit model but OLS has been used by some authors: http://pascal.iseg.utl.pt/~depeco/wp/wp332006.pdf

The problem with Tobit is that it is not yet implemented in statsmodels, so it would be preferable to use OLS for now.

Discussion of how this should really be done (very technical): http://www.clemson.edu/economics/faculty/wilson/courses/bcn/papers/Simar-Wilson-2sdea.pdf

Example of using OLS: http://www.researchgate.net/publication/13645982_Technical_efficiency_in_the_clinical_management_of_critically_ill_patients/links/0deec52b9c71517024000000