Data Mining Quiz Logistic Regression

Logistic Regression

Extends idea of linear regression to situation where outcome variable is categorical

When is Logistic Regression Used

Particularly where a structured model is useful to explain (profiling) or to predict

Logistic Regression Classification

Binary Classification

Logit Goal

Find a function of the predictor variables that relate them to a binary outcome

Logit Defintion

Use a function of y

How can logit be modeled

As a linear function of the predictors

Can a logit be mapped

Yes it can be mapped as a probability and class

Logistic Response Function P

equals the probability of belonging to class one

P size

Has to be between 0 and 1S

Standard Linear Function Q

The number of predictors

What is used to fix problem

Logistic response function

Odds formula

p/ 1 - p

Probability of event

p = odds / 1 + odds

Logit predictors are

infinity to infifty

Popular choice for cutoff

.5

Maximum likelihood estimation

Estimates of B’s are derived through an iterative process

Null hypothesis coefficient for predictors

0