I need to understand and, ideally, resolve the small discrepancy between (unpenalized) Logistic regression results in SAS and Python. As far as I can tell, it does not seem directly attributable to parameter arguments that can be easily changed in either implementation. I'm also curious about the same difference with R, but that's less important and solving Python will probably indirectly answer the same question with R.   I began with a Python-centric approach on StackOverflow, using a semi-famous SAS example from UCLA Stats department. No one there was able to give an answer, so a couple of days ago I opened a bounty on the question and today I thought I'd ping you SAS gurus.   Here's the example on StackOverflow:   https://stackoverflow.com/questions/67128818/source-of-small-deviance-between-sas-proc-logistic-and-python-sklearn-logisticre   Here's a transcription in case you don't want to click the link:   I'm trying to match the results of SAS's  PROC LOGISTIC  with  sklearn  in Python 3. SAS uses unpenalized regression, which I can achieve in  sklearn.linear_model.LogisticRegression  with the option  C = 1e9  or  penalty='none' . That should be the end of the story, but I still notice a small difference when I use a public data set from UCLA and try to replicate their multiple regression of  FEMALE  and  MATH  on  hiwrite . This is my Python script: # module imports import pandas as pd from sklearn.linear_model import LogisticRegression # read in the data df = pd.read_sas("~/Downloads/hsb2-4.sas7bdat") # FE df["hiwrite"] = df["write"] >= 52 print("\n\n") print("Multiple Regression of Female and Math on hiwrite:") feature_cols = ['female','math'] y=df["hiwrite"] X=df[feature_cols] # sklearn output model = LogisticRegression(fit_intercept = True, C = 1e9) mdl = model.fit(X, y) print(mdl.intercept_) print(mdl.coef_) which yields: Multiple Regression of Female and Math on hiwrite: [-10.36619688] [[1.63062846 0.1978864 ]] UCLA has this result from SAS: Analysis of Maximum Likelihood Estimates Standard Wald Parameter DF Estimate Error Chi-Square Pr > ChiSq Intercept 1 -10.3651 1.5535 44.5153 <.0001 FEMALE 1 1.6304 0.4052 16.1922 <.0001 MATH 1 0.1979 0.0293 45.5559 <.0001 which is close, but as you can see the intercept parameter estimate is different at the 3rd decimal place and the estimate on  female  is different at the 4th decimal place. I tried changing some of the other parameters (like  tol  and  max_iter  as well as the  solver ) but it did not change the results. I also tried the  Logit  in  statsmodel.api  - it matches  sklearn , not SAS. R matches Python on the intercept and first coefficient, but is slightly different from both SAS and Python on the second coefficient... Update: I have found that I can affect the estimates by playing with  tol  and also by changing the solver to  liblinear  while removing the penalty in a hack-ish way via the  C  parameter, however no matter how extreme the values it still doesn't quite match SAS. Any thoughts on the source of the error and how to make Python match SAS?   ... View more