Abstract
Model calibration on the 30% UK Biobank testing (validation) set. Calibration was performed on absolute risk final model including all variables and using the built-in calibration from iCARE on a relative risk scale. The observed relative risk in each risk decile is defined as the ratio of the observed proportion of cases in that category and the observed proportion of cases in the validation study (testing set). For comparison, the predicted relative risk is computed as the average of the predicted absolute risk over the subjects in that category and the overall average of the predicted absolute risks in the validation study.