Lecture 6: Logistic Regression (Alternative) Flashcards
(35 cards)
What type of outcome does linear regression model?
Continuous outcome
What type of outcome does logistic regression model?
Categorical outcome
What assumption does linear regression make about relationships?
Linear relationship
Can logistic regression handle non-linear relationships?
Yes
Does logistic regression require normally distributed predictors?
No
What function does logistic regression use to model probabilities?
Logistic regression uses the logistic (or sigmoid) function to model probabilities.
This transforms the linear combination of predictors into a value between 0 and 1, which can be interpreted as a probability.
- It ensures valid probability predictions.
- It models the log odds of the outcome as a linear function of the predictors.
What is the range of values produced by the logistic function?
0 to 1
What must outcome categories be in logistic regression?
Mutually exclusive and exhaustive
What happens if outcomes are not mutually exclusive or exhaustive?
Model assumptions are violated
What are logistic regression coefficients initially expressed as?
Log odds
What are odds in logistic regression?
Ratio of event likelihood to non-event likelihood
What does exponentiating a coefficient b yield?
Odds ratio (e.g., if exp(B) = 1.5 → each extra date increases odds of being in a relationship by 50%)
What does exp(b) = 1.5 mean?
Odds increase by 50% per unit increase in predictor
Are changes in probability constant in logistic regression?
No
Where is the probability change largest in a logistic curve?
Near 0.5
What does b = 0 mean in log odds interpretation?
No effect of predictor
What does exp(b) = 1 mean in odds ratio interpretation?
No effect of predictor
What is the null hypothesis in logistic regression?
Predictor has no effect on the outcome
What test checks if a coefficient significantly differs from 0?
Wald Test
What type of distribution is used in the Wald Test?
Chi-square distribution
What does it mean if the CI for exp(b) excludes 1?
Predictor is significant
What does it mean if the CI for exp(b) includes 1?
Predictor is not significant
What is a pseudo-R² used for in logistic regression?
Assessing model fit
Name two types of pseudo-R².
Cox & Snell, Nagelkerke