Calibration and verification of risk algorithms using logistic regression

The use of logistic regression is proposed as a method of verifying and calibrating disease risk algorithms. The logistic regression model calculates the log of the odds of a binary outcome as a function of a linear combination of predictors. The resulting model assumes a multiplicative (relative) relationship between the different risk factors. Computer programs for performing logistic regression produce both estimates and standard errors, thus permitting the evaluation of the importance of different predictive variables. The use of receiver operating characteristic (ROC) curves is also proposed as a means of comparing different algorithms. An example is presented using data on Sclerotinia stem rot in oil seed rape, caused by Sclerotinia sclerotiorum.