Table 2 Results of simulation studies showing mean values of theta and misclassification rate across 5000 simulations under a range of data-generating scenarios
Scenario | Sample size | Parameter | Method | ||
|---|---|---|---|---|---|
Regression | Kraemer | Modified Kraemer | |||
1a. Simple linear | 50 | θ (SD) | 0.72 (0.676) | 0.46 (0.810) | 0.62 (0.679) |
Misclass. rate | 0.05 | 0.21 | 0.14 | ||
200 | θ (SD) | 0.71 (0.473) | 0.58 (0.555) | 0.68 (0.468) | |
Misclass. rate | 0.04 | 0.15 | 0.09 | ||
300 | θ (SD) | 0.72 (0.337) | 0.65 (0.397) | 0.70 (0.335) | |
Misclass. rate | 0.02 | 0.10 | 0.06 | ||
1b. Simple linear with correlated biomarkers | 50 | θ (SD) | 0.73 (0.629) | 0.50 (0.695) | 0.59 (0.616) |
Misclass. rate | 0.05 | 0.21 | 0.17 | ||
200 | θ (SD) | 0.61 (0.422) | 0.48 (0.483) | 0.56 (0.423) | |
Misclass. rate | 0.05 | 0.17 | 0.11 | ||
300 | θ (SD) | 0.62 (0.276) | 0.57 (0.331) | 0.61 (0.278) | |
Misclass. rate | 0.02 | 0.1 | 0.06 | ||
2. Linear with weak moderators | 50 | θ (SD) | 0.71 (0.665) | 0.45 (0.799) | 0.61 (0.667) |
Misclass. rate | 0.05 | 0.21 | 0.13 | ||
200 | θ (SD) | 0.71 (0.482) | 0.56 (0.567) | 0.66 (0.481) | |
Misclass. rate | 0.04 | 0.15 | 0.09 | ||
300 | θ (SD) | 0.72 (0.336) | 0.66 (0.401) | 0.71 (0.336) | |
Misclass. rate | 0.02 | 0.1 | 0.06 | ||
3. Strong unobserved prognostic marker | 50 | θ (SD) | − 0.23 (3.150) | − 0.24 (3.149) | − 0.31 (3.115) |
Misclass. rate | 0.41 | 0.41 | 0.42 | ||
200 | θ (SD) | − 0.23 (3.150) | − 0.24 (3.149) | − 0.31 (3.115) | |
Misclass. rate | 0.41 | 0.41 | 0.42 | ||
300 | θ (SD) | 0.07 (1.556) | 0.06 (1.570) | 0.01 (1.535) | |
Misclass. rate | 0.32 | 0.32 | 0.34 | ||
4. Strong unobserved moderator variables | 50 | θ (SD) | 0.22 (1.328) | 0.16 (1.405) | 0.14 (1.323) |
Misclass. rate | 0.42 | 0.45 | 0.45 | ||
200 | θ (SD) | 0.43 (0.893) | 0.35 (0.946) | 0.35 (0.911) | |
Misclass. rate | 0.42 | 0.43 | 0.44 | ||
300 | θ (SD) | 0.56 (0.606) | 0.51 (0.654) | 0.52 (0.620) | |
Misclass. rate | 0.41 | 0.42 | 0.42 | ||
5. Misspecified prognostic part | 50 | θ (SD) | 0.53 (0.866) | 0.34 (1.010) | 0.40 (0.882) |
Misclass. rate | 0.17 | 0.25 | 0.22 | ||
200 | θ (SD) | 0.62 (0.582) | 0.49 (0.682) | 0.57 (0.588) | |
Misclass. rate | 0.12 | 0.18 | 0.15 | ||
300 | θ (SD) | 0.67 (0.406) | 0.60 (0.495) | 0.65 (0.407) | |
Misclass. rate | 0.09 | 0.13 | 0.11 | ||
6. Misspecified moderator part | 50 | θ (SD) | 0.70 (0.773) | 0.44 (0.872) | 0.58 (0.775) |
Misclass. rate | 0.16 | 0.25 | 0.21 | ||
200 | θ (SD) | 0.78 (0.559) | 0.60 (0.651) | 0.72 (0.558) | |
Misclass. rate | 0.13 | 0.2 | 0.16 | ||
300 | θ (SD) | 0.80 (0.381) | 0.72 (0.443) | 0.77 (0.382) | |
Misclass. rate | 0.12 | 0.16 | 0.14 | ||
7. Non-linear model | 50 | θ (SD) | 2.11 (8.414) | 1.94 (8.606) | 2.10 (8.566) |
Misclass. rate | 0.45 | 0.45 | 0.44 | ||
200 | θ (SD) | 2.89 (9.242) | 2.84 (10.822) | 1.62 (10.261) | |
Misclass. rate | 0.44 | 0.44 | 0.43 | ||
300 | θ (SD) | 0.31 (10.085) | 0.06 (10.831) | − 0.30 (10.797) | |
Misclass. rate | 0.44 | 0.44 | 0.43 | ||