|
Topics and Objectives:
Hopefully you all recall regression. Regression is quite useful for predicting a dependent variable - as long as it is a numeric variable such
as sales or profits.
This module will introduce you to logistic regression which is used when what you are trying to predict takes on only two discrete values. When dealing with individual customers - we are often interested in predicting whether each customer will respond to the offer or not, whether they are a good credit risk or a bad credit risk, or whether they will renew their cellphone contract or not. For these situations, logistic regression is appropriate. Unfortunately, logistic regression is both more complicated to compute and more complicated to interpret. Modern statistical software packages have removed the computational barriers. So this module will focus on the use and interpretation of logistic regression.
Required Readings:
Applied Logistic Regression
This note walks you through logistic regression using SPSS with an emphasis on how to interpret the results (a little tricky - but doable!)
Assignment 3: Predicting Response at BookBinders: Logistic
Regression
This assignment returns to the BookBinders Book Club data (used in the note of RFM Segmentation) to see how well logistic regression can predict
which customers will respond. The note on RFM Analysis found that RFM was a big improvement over mass mailing. Any bets on whether RFM or logistic regression will 'win'?
Back to top
|
