Table 1 The observed Lp(A)-MI association is magnified by temporal bias.
From: Temporal bias in case-control design: preventing reliable predictions of the future
Initial case state imputation method | Trajectory type | Effect size relative to reported baseline (95% CI) |
|---|---|---|
Weighted sampling | Linear | 0.172 (0.160–0.196) |
Weighted sampling | Logistic | 0.169 (0.150–0.187) |
Weighted sampling | Logarithmic | 0.403 (0.390–0.417) |
Percentile matching | Linear | 0.518 (0.507–0.528) |
Percentile matching | Logistic | 0.517 (0.506–0.527) |
Percentile matching | Logarithmic | 0.639 (0.631–0.647) |
Percent shift | Linear | 0.389 (0.376–0.401) |
Percent shift | Logistic | 0.386 (0.373–0.399) |
Percent shift | Logarithmic | 0.539 (0.530–0.549) |
Weighted sampling | HSF: Impulse in first 10% of trajectory | 0.808 (0.801–0.814) |
Weighted sampling | HSF: Impulse in first 1% of trajectory | 0.980 (0.977–0.984) |
Weighted sampling | HSF: Impulse in first 0.1% of trajectory | 0.998 (0.997–0.999) |
Percentile matching | HSF: Impulse in first 10% of trajectory | 0.898 (0.895–0.901) |
Percentile matching | HSF: Impulse in first 1% of trajectory | 0.989 (0.989–0.989) |
Percentile matching | HSF: Impulse in first 0.1% of trajectory | 0.999 (0.999–0.999) |
Percent shift | HSF: Impulse in first 10% of trajectory | 0.860 (0.857–0.864) |
Percent shift | HSF: Impulse in first 1% of trajectory | 0.987 (0.985–0.989) |
Percent shift | HSF: Impulse in first 0.1% of trajectory | 0.999 (0.999–0.999) |
