Skip to main content
Fig. 2 | Diagnostic and Prognostic Research

Fig. 2

From: Leave-one-out cross-validation, penalization, and differential bias of some prediction model performance measures—a simulation study

Fig. 2

Illustration of the calculation of the c-statistic in leave-one-out cross-validation for the maximum likelihood and ridge estimators. Data consisted of 20 observations of a normally distributed explanatory variable, with 5 randomly chosen observations labeled as “events” (t-test p-value = 0.584). Each tick on the x-axis corresponds to one of the 20 iterations in leave-one-out cross-validation. Grey symbols mark estimated probabilities for the 19 observations used in the model fitting; black symbols mark the estimated probabilities for the left-out observations. Black crosses or circles indicate that the left-out observation corresponds to an event or non-event, respectively. The leave-one-out cross-validated c-statistic was equal to 0.17 for maximum likelihood estimation and equal to 0 for ridge regression

Back to article page