Table 3.
Treatment Effect Estimates by Scenario Using (1) Regression and (2) Differences between the Cure Rate for the Treated and Untreated Patients from Propensity Score Algorithms
| I | II | III | IV | V | ||||
|---|---|---|---|---|---|---|---|---|
| Matching Tolerance Based on Propensity Score¶ | ||||||||
| Scenario | Truth | No Control* | Regression Estimate Controlling for XM† | Inverse Probability Weighting‡ | Average across Bins§ | .1 | .01 | .001 |
| 1 | .200 | .263 | .254 | .255 | .255 | .263 | .258 | .257 |
| 2 | .200 | .273 | .257 | .257 | .257 | .268 | .261 | .260 |
| 3 | .200 | .271 | .260 | .286 | .265 | .265 | .262 | .261 |
| 4 | .200 | .275 | .254 | .255 | .256 | .261 | .254 | .253 |
| 5 | .200 | .282 | .232 | .256 | .239 | .240 | .234 | .233 |
| 6 | .200 | .304 | .227 | .291 | .239 | .242 | .231 | .233 |
| 7 | .200 | .318 | .203 | .299 | .223 | .220 | .212 | .210 |
| 8 | .200 | .338 | .194 | .316 | .220 | .213 | .222 | .225 |
*
Difference in cure rates between treated and untreated patients without control for XM.
†
Linear multiple regression estimate of the effect of T on cure controlling for XM.
‡
Weighted difference in cure rates between treated and untreated patients with weights based on the propensity score.
§
Average of the difference in cure rates between treated and untreated patients across the five propensity score bins.
¶
Difference in cure rates between propensity score-matched treated and untreated patients.