Logistic Regression

Useful information regarding the Logistic Regression method

First, see the sub-page "Why Logistic vs. Linear Regression".

NIST gives cursory mention of logistic regression here (under the paragraph titled "Numerical Methods' Forte")

Good overviews on Logistic Regression and implementation, include some (not all) that use the R software package:

• [good place to start] By Dr. David Garson at North Carolina State University http://faculty.chass.ncsu.edu/garson/PA765/logistic.htm

• from Stanford (Christopher Manning): http://nlp.stanford.edu/manning/courses/ling289/logistic.pdf

• from UCLA: http://www.ats.ucla.edu/stat/r/dae/logit.htm

• Also see the PowerPoint presentation file "Math456FinalProject-LogisticRegression.ppt" attached at the bottom of this page.

Related methods and concepts

• Polytomous Logistic Regression

  • Multinomial models

  • Ordered (cumulative) logit models

• Log-linear analysis (see Advanced Log-Linear Models Using SAS (Zelterman))

• Crosstab tables and chi-square analysis

• Odds Ratios

• Probit and Tobit regression

• Poisson regression

• Sensitivity and Specificity

• ROC Curves

Interpreting Odds Ratios in Logistic Regression

• UCLA idre - http://www.ats.ucla.edu/stat/mult_pkg/faq/general/odds_ratio.htm

Software packages

Statistica GLZ module

SAS has several packages that will do logistic regression.

• proc CATMOD

• proc GENMOD

• proc LOGISTIC

• (Firth)

• other?

R has several packages as well:

• stat:glm

• Design:lrm

• Hmisc

• rms

• logistf

• elrm (exact methods for lrm)

• Google search "Logistic Regression using R"

• (there are others that don't come to mind as this entry is made ... will add later if useful)

• General R note: "plogis" has consequently been called the ‘inverse logit’. Check out ?plogis

Many other packages as well (Minitab, S-Plus, etc. etc.)

Good references include

Applied Linear Regression Models (Kutner)

Categorical Data Analysis Using the SAS System (Stokes)

Introduction to Categorical Data Analysis (Agresti)

Logistic Regression Using the SAS System (Allison)

SAGE Logistic Regression - A Primer (Pampel)

SAGE Interaction Effects in Logistic Regression (Jaccard)

Other related references include

A Guide to Modern Econometrics (Verbeek)

Advanced Log-Linear Models Using SAS (Zelterman)

Data Analysis and Graphics Using R (Maindonald)

Modern Applied Statistics with S-Plus (Venebles)

Logistic Regression course notes (DH)

Forms of the response variable

• single-trial syntax, is applicable to binary, ordinal, and nominal response data.

• events/trials syntax, is restricted to the case of binary response data.