Fig. 4From: Impact of sample size on the stability of risk scores from clinical prediction models: a case study in cardiovascular diseasePercentile ranges of risk for individuals, subsetted by calibration-in-the-large of the models (female cohort, N = 10,000). Each data point represents the 5–95th percentile range in risk for an individual across a group models defined by their calibration-in-the-large. Two thirds of models had calibration-in-the-large within 0.25% of the population-derived model, one third of models had calibration-in-the-large within 0.14% of the population-derived modelBack to article page