Module 10 Summary

What this module established

A probability cannot be negative, and a straight line has no way of knowing that — smf.ols on a churn indicator predicted −0.088 and −0.510. Logistic regression models the unbounded log-odds instead.

Carry forward

smf.ols and smf.logit take the same formula and only one is right. The wrong one returns a full summary with no error.
Exponentiated coefficients are odds ratios. With a churn rate of 0.497 an odds ratio of 2.45 badly overstates the risk ratio.
The same six-month step moved the probability by 0.067, then 0.116, then 0.061 — steepest near p = 0.5. One coefficient is not a change in probability.
On a 5.9% positive rate the model scored 0.9415 accuracy while catching 0 of 117 positives — identical to a rule that always says no.

Before moving on

Move on when your report gives predicted probabilities rather than coefficients, and your threshold has a stated justification.