Labs

Browser labs for the course’s hardest ideas. Everything runs in your browser; no installs, no server. Each lab follows one loop: predict before moving a control, manipulate one quantity at a time, observe the exact consequence, explain it in ordinary language, and transfer an object to Stata for the problem set.

Every output says whether it is a live calculation, a stored simulation result, or a conceptual illustration. Illustrative bands are never presented as estimated uncertainty.

Lab	Misconception it exposes	Lecture	Status
Shock-to-Response Explorer	That a response is estimated from one episode	1	in development
Transformation Workbench	That an accumulated change is an accumulated level	2	in development
Identification and Controls Sandbox	That “I included controls” is an identification argument	3	in development
Instrument and Multiplier Laboratory	That a strong first stage establishes causality	4	in development
Confidence-Interval Coverage Simulator	That a narrower interval is a more reliable one	5	in development
Claims About a Curve	That “significant here, insignificant there” tests a difference	6	in development
Bias–Variance Laboratory	That a smoother curve is a better estimate	7	in development
Smoothing Workbench	That the fitted shape came from the data	8	in development
State-Dependence Laboratory	That an interaction identifies a fixed-state experiment	9	in development
Causal-Weight Explorer	That a moved coefficient means a moved response function	10	in development
Where Does Identification Come From?	That more rows mean more identifying information	11	in development
Clean-Control Builder	That already-treated units are valid controls	12	in development
Counterfactual and Sensitivity Workbench	That a reproduced calculation is a policy experiment	13	in development
Replication Referee Desk	That a matching figure means a sound analysis	14	in development