Chapter Twelve: Logistic Regression

1
Chapter Twelve Logistic Regression

• DBA 903
• Research Methods III
• J. Schreiber

2
Why Logistic Regression?

• Allows one to predict ________________ from a set
  of variables.
• Variables may be ___________________.
• Makes no assumption about ____________________.
• Cannot produce negative probabilities.
• Outcomes

3
The Outcome Variable (Y-hat)

• The probability of having a particular outcome.
• The logit (log of the odds) is a linear
  combination of predictors.
• Goal Find best linear combination of predictors
  to maximize the likelihood of reproducing
  observed outcome frequencies.

4
Kinds of Research Questions

• Comparing Models
• Importance of Predictors

5
Kinds of Research Questions

• Parameter Estimates
• Classifying Cases
• Interactions
• Covariates
• Strength of Association

6
Limitations

• Theoretical
• Causal
• Other methods are more powerful
• Practical
• Sample size
• Adequacy of Expected Frequencies

7
Limitations Practical (continued)

• Linearity in the Logit
• Multicollinearity
• Outliers
• Independence of Errors

8
Important Issues

• Types of Logistic Regression
• Direct, Sequential, Stepwise
• Probit
• Assessing Goodness-of-Fit of Models
• Testing Individual Variables

9
Important Issues (continued)

• Number and Type of Outcome Categories
• Strength of Association for a Model
• Coding Outcome and Predictor Categories

10
Important Issues (Continued)

• Classification of Cases
• Interpretation of Coefficients using Odds
• Importance of Predictors