Logistic Regression

Question 1

Logistic Regression

Answer 1

Predicting a categorical outcome from continuous and/or categorical predictors. Can be binary (2 levels of DV) or multinomial (3+ levels of DV). Analyses the log odds of an event occurring.

Question 2

Logistic Regression in SPSS

Answer 2

Block 0 - initial model with no predictors.
Block 1 - including one or more predictors.
Iteration History - has the chi square value
Model summary gives pseudo R square values
-2loglikelihood (deviance) is an indicator of how much unexplained variance is in the model (big is bad)
Variables in the Equation - gives Exp(B), which is the odds ratio

Question 3

Sensitivity

Answer 3

How well predictors actually predict a positive outcome

Question 4

Specificity

Answer 4

How well predictors actually predict a negative outcome

Question 5

False Positive (False Alarm)

Answer 5

How often a positive outcome is predicted but did not occur.

Question 6

False Negative (Miss Rate)

Answer 6

How often a negative outcome is predicted but did not occur.

Question 7

Assumptions of Logistic Regression

Answer 7

Linearity of the logit, homogeneity of variance, normality, independence, expected counts (chi square used to calculate deviance), multicollinearity.

Question 8

Complete Separation

Answer 8

Predictor perfectly predicts the outcome. Produces huge standard errors.

Question 9

Incomplete Information

Answer 9

One or more combinations of variables are not represented.

Question 10

Overdispersion

Answer 10

Observed variance is bigger than predicted by the model. Caused by violations of independence.