What is the equivalent Julia code for statsmodels.api.OLS

[Shayan]

@sgaure

Just to be complete. Python uses a pseudo inverse for degenerate systems. There is really no canonical way to solve singular problems. You can mimic python’s behaviour as follows:

using LinearAlgebra

Xc = hcat([1,1,1], X)   # add a constant
cvmat = pinv(Xc' * Xc)  # compute covariance matrix as pseudoinverse
coefs = cvmat * Xc' * y   # perform the OLS
serr = sqrt.(diag(cvmat) .* 3/2)   # find standard errors, small sample adjust by N/(N-dof)
coefs[2]/serr[2]   # compute t-score

Thank you so much for your replies. I don’t understand why you’ve used the constant 3/2. Anyway, the result is still different:

Am I missing something here?