Logistic Regression

Question 1

What is the key difference between linear regression and logistic regression?

Answer 1

Multiple linear regression evaluates predictors of continuously distributed outcomes while multiple logistic regression evaluates predictors of dichotomous outcomes, i.e. outcomes that either occurred or did not.

Question 2

Why do you take the log odds of a logistic regression analysis?

Answer 2

When you take a dichotomous outcome that either occurred or did not occur and express it as odds, i.e. as a continuously distributed outcome, you tend to get a curvilinear relationship. Taking the log(odds) will make the relationship fairly linear. You must exponentiate final odds ration to get the final result.

Question 3

Odds of an outcome occurring in logistic regression

Answer 3

log[P/(1-P)]

Question 4

In logistic regression, what kinds of predictor variables can you have?

Answer 4

Independent predictor variables can be a mix of continuous, dichotomous, or dummy variables (ordinal or categorical).