ABSTRACT
Binary outcomes are commonplace in people analytics and many other fields. Promotion and attrition are examples of binary outcomes in the workplace, and often other forms of outcomes are converted to binary through the use of cutoffs. This chapter deals with modeling the likelihood of a binary event using binomial logistic regression. It introduces the logistic function as an approximation of a probability distribution, and proceeds to demonstrate how this gives rise to a highly interpretable model based on the concept of odds. Using an example related to the promotion of salespeople, the chapter goes through the steps of fitting a binomial logistic regression model, measuring its fit with the data, interpreting the coefficients as log odds and converting these to odds ratios. The concepts of model parsimony and simplification are introduced. More complex topics such as determining goodness of fit and performing diagnostics on these models are also touched upon in this chapter, which also sets a foundation for studying other log-likelihood models covered in later chapters.
