Skip to main content
. 2019 May 31;38(18):3444–3459. doi: 10.1002/sim.8183

Table 4.

Effects of differential measurement of predictors in events and nonevents in four scenarios. Mean c‐statistic, median calibration slope, and mean Brier score (standard deviation) averaged over 10 000 repetitions for a single‐predictor logistic regression model under four specific measurement error structures varying in the degree of random measurement variance under the differential measurement error model (Equation 1). By default, σϵ2 is set to 1.0. When σϵ12=0.5, measurements are more precise in cases. When σϵ12=2.0, measurements are less precise in cases

C‐statistic Calibration Brier Score
Differential Measurement Error at… Derivation Validation Slope Derivation Validation
Derivation
σϵ12=0.5
0.730 (0.011) 0.707 (0.012) 0.780 (0.071) 0.209 (0.004) 0.219 (0.004)
σϵ12=2.0
0.655 (0.012) 0.707 (0.012) 1.856 (0.208) 0.231 (0.003) 0.223 (0.002)
Validation
σϵ12=0.5
0.706 (0.012) 0.730 (0.011) 1.293 (0.120) 0.217 (0.003) 0.211 (0.003)
σϵ12=2.0
0.706 (0.012) 0.655 (0.012) 0.547 (0.061) 0.217 (0.004) 0.237 (0.005)