The plan, in three passes

The final is all multiple choice, formatted like the two midterm practices, and weighted to what came after them. Spend your time in that order.

Pass 1: Own the post-midterm material (roughly 60% of your time)

Lectures 12 to 17: regression discontinuity, difference-in-differences, assessing studies and mechanisms, turning statistics into substantive answers, measurement, and the limits of quantification.

• For each lecture: read the guide (10 min), then take that lecture's drill (15 min), then review every explanation, including the ones you got right.
• DiD: always draw the 2x2 table. RDD: always ask what the running variable and cutoff are, and who is just on either side.
• These two designs plus mechanisms questions are where a scenario block can hide 5 or 6 easy marks.

Pass 2: Rebuild the midterm base (roughly 25%)

• Retake both midterm practice exams cold, then check with the answer keys.
• Drill any lecture (2 to 11) where you dropped more than two marks: correlation vs causation, ATT + Bias + Noise, confounders vs mechanisms (control the confounder, never the mechanism), CIs, p-values, sample size logic, publication bias.
• Skim the formula sheet page until every formula has a "when do I use this" attached in your head.

Pass 3: Simulate (the last stretch)

• Take Practice Final A in one timed sitting. Review it fully. A day later, take Practice Final B.
• For every miss: identify which lecture it maps to, reread that guide section, redo that drill.
• The exam rewards careful reading. Half the distractors are true statements that answer a different question.

Exam-day habits

• Read the scenario once slowly before touching its questions; the same setup feeds 4 to 8 of them.
• Interpretation questions: pick the answer with the caveats ("on average", "associated with", "if the null were true"). Absolute words like "guarantees" and "proves" mark wrong options.
• R code questions: lm(y ~ x + control) adjusts for a control, y ~ x * z adds an interaction, geom_smooth(method = "lm") draws regression lines.