Scribd
Upload
222 views

Logistic Regression

This document provides an introduction to logistic regression, including: - Logistic regression is used to model binary outcome variables and predict the probability of occurrence based on independent variables. - It allows for both continuous and categorical predictor variables, and does not require normality assumptions like discriminant analysis. - The logistic model transforms the probability using the logit function (the log of the odds ratio) to make it suitable for regression. - SAS procedures like LOGISTIC and GENMOD can be used to perform logistic regression, and options allow specification of variables, model selection, and evaluation of model fit and predictive accuracy.

Uploaded by

Nikhil Gandhi
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
222 views

Logistic Regression

This document provides an introduction to logistic regression, including: - Logistic regression is used to model binary outcome variables and predict the probability of occurrence based on independent variables. - It allows for both continuous and categorical predictor variables, and does not require normality assumptions like discriminant analysis. - The logistic model transforms the probability using the logit function (the log of the odds ratio) to make it suitable for regression. - SAS procedures like LOGISTIC and GENMOD can be used to perform logistic regression, and options allow specification of variables, model selection, and evaluation of model fit and predictive accuracy.

Uploaded by

Nikhil Gandhi
Copyright
© © All Rights Reserved
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 10

LOGISTIC REGRESSION

1
C
O
M
P
-
S
T
A
T

G
R
O
U
P

--Models with binary outcome
CONTENTS

Introduction
The Logistic model
An example
SAS code
Results and analysis

2
C
O
M
P
-
S
T
A
T

G
R
O
U
P

INTRODUCTION
Situations where outcome is a qualitative binary variable. for
example true or false, on or off ,success or failure etc.

We want to model our response on a set of dependant
variables(continuous or discrete).

Why logistic regression?
A multiple linear regression model deals with a quantitative
response only.

Predicts the probability of occurrence of a particular response.
for example probability of success/failure.
You want to predict whether a loan applicant will default
or not based on certain attributes provided







3
C
O
M
P
-
S
T
A
T

G
R
O
U
P

Advantages over discriminant analysis
No need of multivariate normality assumption
Can include categorical as well as continuous predictor
variables
Discriminant analysis better when all assumptions are
satisfied

Like linear regression concerns are same to check the
goodness of fit ,explanatory capability of predictor
varaiables,issue of multicollinearity, looking for
outliers, predicting new observations
4
C
O
M
P
-
S
T
A
T

G
R
O
U
P

THE LOGISTIC MODEL
5
C
O
M
P
-
S
T
A
T

G
R
O
U
P

Not that the model is nonlinear in p, the
probability of occurrence of the event.

The term L= (p/1-p) is called ODDS
RATIO,i.e. the odd in favor of
occurrence of the event.

Log of the odds ratio is called the logit.



6
C
O
M
P
-
S
T
A
T

G
R
O
U
P

Consider a study on cancer remission. The data,
consisting of patient characteristics and whether or
not cancer remission occurred
AN EXAMPLE
LOGISTIC REGRESSION IN SAS
SAS provides four procedures to perform logistic regression
PROC LOGISTIC
PROC GENMOD
PROC PROBIT
PROC CATMOD (multinomial logistic regression)
Each procedures has its own advantages and applicability
7
C
O
M
P
-
S
T
A
T

G
R
O
U
P

Running logistic regression
Proc logistic data=test descending;
/*default is to estimate a model predicting the lowest
value of the dependent variable*/
model y=x;run;
proc genmod data=;
model y=x/dist=binomial;run;
Class option
Allow to specify that a variable is to be treated as
categorical
Precedes the model statement






8
C
O
M
P
-
S
T
A
T

G
R
O
U
P

THE MODEL STATEMENT IN PROC LOGISTIC
Some important options
SELECTION= specifies effect selection method
LACKFIT requests Hosmer and Lemeshow goodness-of-fit test
Options for ROC Curves
OUTROC= names the output data set
ROCEPS= specifies probability grouping criterion
INFLUENCE displays influence statistics
RSQUARE displays generalized R2
CTABLE displays classification table
PEVENT= specifies prior event probabilities
PPROB= specifies probability cut points for classification
9
C
O
M
P
-
S
T
A
T

G
R
O
U
P

Sensitivity
Percentage of correctly predicted EVENT responses
Specificity
Percentage of correctly predicted NON EVENT responses
Use outroc= option in model statement
A plot of the ROC curve can be constructed by using the PLOT or
GPLOT procedure with the OUTROC= data set and plotting
sensitivity (SENSIT) against 1-specificity (1MSPEC)

10
C
O
M
P
-
S
T
A
T

G
R
O
U
P

You might also like