Difference in Python statsmodels OLS and R's lm

Question

I\'m not sure why I\'m getting slightly different results for a simple OLS, depending on whether I go through panda\'s experimental rpy interface to do the regression in

Answer 1

Looks like Python does not add an intercept by default to your expression, whereas R does when you use the formula interface..

This means you did fit two different models. Try

lm( y ~ x - 1, data)

in R to exclude the intercept, or in your case and with somewhat more standard notation

lm(num_rx ~ ridageyr - 1, data=demoq)

Answer 2

Note that you can still use ols from statsmodels.formula.api:

from statsmodels.formula.api import ols

results = ols('num_rx ~ ridageyr', demoq).fit()
results.summary()

I think it uses patsy in the backend to translate the formula expression, and intercept is added automatically.