Table I.
Simulated power for detecting a between-group difference using conditional logistic regression with stratification based on propensity scores (sample size per group = 1000, significance level: 0.05)
| Variable | Number of strata | Group 1 | Group 2 | Number of strata | Group 1 | Group 2 | |
|---|---|---|---|---|---|---|---|
| Scenario 1 | Scenario 2 | ||||||
| Distribution of PS | Beta(2,4) | Beta(7,1) | Beta(2,4) | Beta(7,1) | |||
| Binomial proportion for success | 0.1 | 0.1 | 0.1 + PS/2 | 0.1 + PS/2 | |||
| 5 10 15 20 30 |
0.05 0.05 0.05 0.05 0.05 |
5 10 15 20 30 |
0.21 0.06 0.05 0.05 0.05 |
||||
| Scenario 3 | Scenario 4 | ||||||
| Distribution of PS | Beta(4,2) | Beta(2,4) | Beta(4,2) | Beta(2,4) | |||
| Binomial proportion for success | 0.1 + PS/2 | 0.19 + PS/2 | 0.19 + PS/2 | 0.1 + PS/2 | |||
| 5 10 15 20 30 |
0.76 0.83 0.84 0.84 0.84 |
5 10 15 20 30 |
0.93 0.88 0.87 0.86 0.86 |
||||
| Scenario 5 | Scenario 6 | ||||||
| Distribution of PS | Beta(4,2) | Beta(2,4) | Beta(4,2) | Beta(2,4) | |||
| Binomial proportion for success | 0.1 + PS/8 | 0.1 | 0.1 | 0.1 + PS/8 | |||
| 5 10 15 20 30 |
0.87 0.85 0.85 0.84 0.84 |
5 10 15 20 30 |
0.86 0.86 0.86 0.86 0.85 |
||||
| Scenario 7 | Scenario 8 | ||||||
| Distribution of PS | Beta(2,4) | Beta(7,1) | Beta(4,2) | Beta(7,1) | |||
| Binomial proportion for success | 0.3 + PS/2 | 0.1 + PS/2 | 0.1 + PS/2 | 0.18 + PS/2 | |||
| 5 10 15 20 30 |
0.88 0.92 0.93 0.94 0.94 |
5 10 15 20 30 |
0.89 0.84 0.83 0.82 0.82 |
||||
| Scenario 9 | Scenario 10 | ||||||
| Distribution of PS | Beta(4,2) | Beta(7,1) | Beta(4,2) | Beta(2,4) | |||
| Binomial proportion for success | 0.1 | 0.15 | 0.1 | 0.15 | |||
| 5 10 15 20 30 |
0.78 0.77 0.77 0.77 0.76 |
5 10 15 20 30 |
0.70 0.69 0.68 0.68 0.68 |
||||
PS – propensity score, Beta(α,β): beta distribution with parameters α and β.