Logistic Regression using One Predictor
- Introduction and Links to Materials
- PROC LOGISTIC
- Videos (49:56)
In Unit 4 we will review two-variable methods for binary outcomes along with the measures of association: excess risk, relative risk, and odds ratio. Then we will provide an introduction to logistic regression using one predictor in each of the cases we discussed for linear regression but we will cover them in a different order: one binary predictor, one multi-level predictor, and one continuous predictor. Then in Unit 5 we will look at including multiple predictors including interactions as we did for multiple linear regression.
The right side of our models will look almost identical to that for linear regression and many of the techniques and ideas translate easily from linear to logistic regression. However, the added mathematical details introduced with logistic regression often require extra attention. We need to remember rules relating to LOGARITHMS and EXPONENTIAL functions.
Please carefully review the videos we provide and go through those examples yourself carefully to help prepare you for assignments. We will also provide full SAS code and output for this unit which will be useful for the software assignments as well as for future reference for those of you planning to conduct regression analysis in your own research.
Consider the following materials from Penn State STAT 501 as support for our lecture materials for this content.
We will be using PROC LOGISTIC for simple and multiple logistic regression so let’s look at some PROC LOGISTIC documentation including examples.
Introduction to Logistic Regression (10:30)
Logistic Regression for One Binary Predictor (9:17)
Logistic Regression with One Multi-Level Categorical Predictor
Part A (8:14)
- Slides 1-16
Part B (9:17)
- Slides 17-end