I am building a 2-variable predictive model and would like to obtain the best predicting point (i.e. Youden Index).
The most ideal predictive probability in my 2-variable model (based on the best balance between sensitivity & specificity in the AUC graph) is 0.335 (pls see the image attached).
My question is: How can I convert this probability back to the Youden Index (i.e. How can I conclude what age or BMI value will have the most utility in predicting the disease)?
Below is my code:
fit<- glm(disease~ Age + BMI, data=mydata, family=binomial) pred.val.3 <- predict(fit, type ="response") auc((mydata$disease), pred.val.3, plot=TRUE, print.thres="best", auc.polygon=TRUE, auc.polygon.col="lightblue", asp=FALSE, print.auc=TRUE, print.auc.cex=2) ci.auc((mydata$disease), pred.val.3)