Figure 3From: Evaluation of prediction models for the staging of prostate cancer Calibration curves which illustrate predicted probabilities on the x-axis and the actual outcome on the y-axis. In this case the outcome is a binary variable hence Loess smoothing (blue line) was used to estimate the actual outcomes for (a) logistic regression model, (b) random forests model, (c) k nearest neighbours model, (d) biopsy Gleason score, (e) PSA and (f) clinical stage.Back to article page