As discussed in the previous section, the goal of logistic regression is to model the probability of a given outcome occurring. However, rather than predicting probabilities, researchers sometimes want the output of their model to indicate if either a success or a failure is expected for a given X value. This is called classification. The simplest way to perform classification is to set what is known as a cutoff value. This value is a number between 0 and 1 that serves as the division for what to call a “success” and what to call a “failure”. For example, setting the classification cutoff to 0.5 is common (and default for simple logistic regression in Prism), and means that if the model predicts a probability of success greater than or equal to 0.5, then that prediction is classified as a "success" (Y=1), while if the model predicts a probability less than 0.5, it will classify it as a "failure" (Y=0).

There are a LOT of metrics that researchers use from this sort of classification including concepts like the sensitivity and specificity of a model, the true positive rate (TPR) and false positive rate (FPR) of classification, the positive and negative predictive power of the model, and much more. Read more about classification.