The model that was too perfect

Do internet access and study hours actually predict academic success — or is that just what everyone assumes?

My regression returned R² = 1.000. In real data, perfection is a symptom, not a triumph. I dug in and found the leak: the “final percentage” was a derived field — an equal-weighted average (33.3% each) of the three subject scores. The model had been predicting the answer with the answer.

Rebuilt on independent predictors, the story flipped. Students without internet passed at 94.8% versus 94.6% with it — a 0.2-point gap that makes connectivity a non-factor in this sample. Study hours showed r = −0.013 against final grades: essentially no relationship. Success here was subject mastery, which argues for mastery-focused teaching over time-based study mandates.