Question
Per a model named for this statistician, the fundamental problem of causal inference states that it is impossible to observe the effect of more than one treatment on a subject and to directly observe causal effects. For 10 points each:
[10h] Name this Harvard statistician whose causal model uses Jerzy Neyman’s (“YAIR-zhih NAY-min’s”) potential outcomes framework and a probabilistic assignment mechanism to estimate missing counterfactuals.
ANSWER: Donald Rubin [or Donald Bruce Rubin]
[10m] When non-compliers exist, these variables can be used to estimate causal relationships. These variables must be correlated with endogenous explanatory variables and are subject to the exclusion restriction.
ANSWER: instrumental variables [or IV]
[10e] Non-compliance can necessitate adjusting for post-treatment covariates by using the stratified form of this technique. This technique involves selecting individuals at random from a population.
ANSWER: sampling [accept stratified sampling]
<Chicago B, Social Science>
Summary
| California | 2025-02-01 | Y | 3 | 6.67 | 67% | 0% | 0% |
| Florida | 2025-02-01 | Y | 3 | 6.67 | 67% | 0% | 0% |
| Lower Mid-Atlantic | 2025-02-01 | Y | 6 | 11.67 | 100% | 17% | 0% |
| Midwest | 2025-02-01 | Y | 6 | 10.00 | 100% | 0% | 0% |
| North | 2025-02-01 | Y | 3 | 10.00 | 100% | 0% | 0% |
| Northeast | 2025-02-01 | Y | 5 | 14.00 | 100% | 20% | 20% |
| Overflow | 2025-02-01 | Y | 5 | 12.00 | 100% | 20% | 0% |
| South Central | 2025-02-01 | Y | 2 | 15.00 | 100% | 50% | 0% |
| Southeast | 2025-02-01 | Y | 4 | 10.00 | 100% | 0% | 0% |
| UK | 2025-02-01 | Y | 10 | 12.00 | 100% | 20% | 0% |
| Upper Mid-Atlantic | 2025-02-01 | Y | 8 | 12.50 | 100% | 25% | 0% |
| Upstate NY | 2025-02-01 | Y | 3 | 13.33 | 100% | 33% | 0% |
Data
| BU | UMass Boston | 0 | 0 | 10 | 10 |
| Harvard A | Brandeis A | 10 | 10 | 10 | 30 |
| Harvard B | Brandeis B | 0 | 0 | 10 | 10 |
| Dartmouth A | Brown A | 0 | 0 | 10 | 10 |
| MIT A | Tufts A | 0 | 0 | 10 | 10 |