Skip to main content
Fig. 5 | Diagnostic and Prognostic Research

Fig. 5

From: Reclassification calibration test for censored survival data: performance and comparison to goodness-of-fit criteria

Fig. 5

Power when an interaction term is omitted for 0, 25, and 50% censoring rates. The RC and HL statistics were calculated for the following reduced and full models: y~ x1 and y~ x1 + x2; y~ x1 and y~ x1 + x12; and y~ x1 + x2 and y~ x1 + x2 + x1 * x2. Data were simulated according to the full model, but the reduced model was used to estimate the expected proportions in the RC statistic formula. Left column—power of the RC-GND test, right column—power of the HL gof test. N = 1000 (top row), 5000 (middle row), 10,000 (bottom row), p = 0.1, collapse when evg < 5

Back to article page