Bayesian Methods for Management and Business: Pragmatic Solutions for Real Problems

Chapter 2A First Look at Bayesian Computation

2.1 Getting Started

In order to put Bayes’ theorem to use, we first have to make three decisions. First, we must select a sampling distribution or likelihood function for the outcome data. Next we must think about what factors might be influencing the outcome data so as to identify parameters and select a particular structure for them. Finally, we must assign a prior distribution for all the parameters we have identified in the previous decision.

The first two decisions are usually typically discussed as one decision, which is called selecting the model by other authors. In this book, we will discuss the first two decisions separately. The ordering in which the decisions is made is not critical as long as all three decisions are made.

Inevitably, these decisions involve considerations about your goals and your data. Investing the time with these considerations ultimately makes the Bayesian approach very powerful because of the flexibility it provides. As an analogy, it is possible to travel between two cities by train, and this may even be the most economical option in a number of circumstances. But learning the different tasks required to operate a car gives the driver many more options.

This chapter will focus on proportion data, an important kind of business data. Examples of this kind of data include the percentage of people who remember an advertisement, or the percentage of people who return a particular item to the store. ...