Table 2. Small sample (simulated) and asymptotic (calculated) relative efficiencies of the estimates of the log-linear interaction parameter. *.
Case-control ratio | Sampling method | Small sample efficiency† |
Asymptotic relative efficiency‡ |
1:1 | Matched | 0.42 | 0.40 |
Counter-matched | 0.43 | 0.43 | |
Unmatched | 0.30 | 0.28 | |
Matched | 0.63 | 0.58 | |
1:2 | Counter-matched: | ||
2 unexposed, 1 exposed | 0.89 | 0.80 | |
1 unexposed, 2 exposed | 0.49 | 0.45 | |
Unmatched | 0.48 | 0.44 | |
1:3 | Matched | 0.74 | 0.68 |
Counter-matched | 0.90 | 0.85 | |
Unmatched | 0.59 | 0.54 | |
1:5 | Matched | 0.85 | 0.78 |
Counter-matched | 0.95 | 0.91 | |
Unmatched | 0.73 | 0.66 | |
1:9 | Matched | 0.95 | 0.87 |
Counter-matched | 0.98 | 0.95 | |
Unmatched | 0.86 | 0.78 |
* Relative efficiencies are given for the model with exp{α} = exp{β} = 2 and exp{γ} = (8/7).
† The small sample efficiency is the ratio SE2( )cohort/SE2( )sample of squared standard errors of the parameter estimates from the simulated data.
‡ The asymptotic relative efficiency is the ratio SE2( )cohort/SE2( )sample of squared asymptotic standard errors.