Deck 12: Propensity Score Analysis

Question 1

Matching that groups treatment and control units into subgroups based on approximately equal propensity scores is which one of the following?

A) one to one
B) one to two
C) one to many
D) stratified

Answer 1

D
Explanation: Stratification involves ranking cases based on their estimated propensity scores and then stratifying into mutually exclusive and equal-sized subgroups.

Question 2

With which one of the following is the propensity score applied in the efficacy analyses?

A) covariance adjustment
B) matching
C) stratification
D) weighting

Answer 2

D
Explanation: Weighting on the propensity score is an additional way that PS can be applied to balance groups (Luellen, Shadish, & Clark, 2005). Should the researcher decide to apply the propensity score as a sampling weight, this step of matching on the propensity score is skipped (as there is no matching entailed), and the researcher proceeds directly to efficacy analyses.

Question 3

Which one of the following is true about the counterfactual?

A) it can be estimated
B) it is observable
C) it is the observed outcome that occurs as a result of the cause
D) the counterfactual for treated cases reflect the outcome when treated

Answer 3

A
Explanation: The Neyman-Rubin counterfactual framework of causality posits that the counterfactual can indeed be estimated.

Question 4

Which one of the following is the first step in propensity score analysis?

A) estimate the predicted probability of treatment
B) compute model adequacy
C) match on the propensity score
D) estimate the treatment efficacy

Question 5

Which step in the propensity score process is balance determined?

A) estimate the predicted probability of treatment
B) compute model adequacy
C) match on the propensity score
D) estimate the treatment efficacy

Question 6

Which one of the following reflects the goal when estimating the propensity score?

A) balance the means and variances
B) evaluate parameter estimations resulting from the propensity score estimation
C) obtain the propensity score
D) transform the units of analysis

Question 7

Comparable distributions of covariates in groups is reflected by which one of the following?

A) balance
B) equal means and variances on the outcome
C) inclusion of polynomials
D) large effects between groups

Question 8

The overlap between the treated and untreated propensity score distributions is reflected by which one of the following?

A) balance
B) large effects between groups
C) region of common support
D) stratification

Question 9

Estimates of causal effects can be applied to the majority of the sample in which one of the following conditions?

A) when there is balance
B) when there is a great deal of overlap between the propensity score distributions of the groups
C) where there are large effects between treated and untreated cases
D) when there are differential distributions of covariates

Question 10

Which one of the following is NOT a decision that is needed for researchers who use matching as the conditioning strategy?

A) algorithm
B) balance
C) distance
D) structure