ABSTRACT
Logistic Regression (LR) is a popular tool for binary classification and considered as an alternative to LDA. While LDA assumes normality of the variables under study, the LR method does not make it and hence applies in more general contexts. Further both continuous and categorical variables can be handled in logistic regression. Predictive models for several events in clinical studies can be developed by using the LR model. One of the interesting problems in clinical research is the problem of inventing new biomarkers and their cutoff values. LR is the popularly used tool in this direction.
This chapter is aimed at training the user on the conceptual ideas as well as practical outlook on conducting LR. The concept of Odds Ratio (OR) which is an important output from LR are discussed in detail with live data. Several statistical software like SPSS, R, SYSTAT, SAS and even Excel (with Real Statistics Add-in) and MedCalc provide modules to perform LR. (157)
