Logistic regression
Predicting dichotomous outcomes is central to epidemiologic science and clinical care and is used to describe and test the relationship between a dichotomyous outcome and one or more potentially predictive variables.
Abstract
L ogistic regression is probably the most frequently used modeling tool in medical research. It is used to describe and test the relationship between a dichotomous outcome and one or more potentially predictive variables. A dichotomous outcome is any mutually exclusive but all-inclusive pair like dead or alive, pass or fail, and yes or no. Predicting dichotomous outcomes is central to epidemiologic science and clinical care. For example, how does the degree of leukoreduction affect the probability of febrile transfusion reactions? What decrease in central venous pressure, controlling for other variables, is associated with avoiding RBC transfusion in partial hepatectomy surgery? These and similar questions are common in medical research, where complex information is reduced to binary decisions.