Logistic regression model assumes
- the true conditional probability is a linear combination of the predictors;
- the observations are independent;
- each observation has an error term thst is normally distributed;
- each observation has an error term that has constant variance
As one decreases the classification threshold p0
- sensitivity will never decrease
- specificity will never decrease
- the sum of sensitivity and specificity stays constant
- the overall misclassification rate stays constant