Maximum Likelihood Estimation
Let y denote the realized values of a set of observations and
P(Y, ) denote the joint density of Y = y and = (1 , . . . , q )
with belonging to some subset U Rq . The Likelihood of
given the observations is dened to be a fun

Condence Intervals for Odds Ratios
The sample odds ratio is:
n11 n22 .
=
n21 n12
When any of the cell counts are zero, equals 0 or .
If both entries in a row or column are zero, is undened.
the variance, then, doesnt exist.
We can amend the estimator by

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
February 2, 2015
Components of a GLM
Generalized linear models (GLMs) : extension of ordina

Welcome to BIST-512
Categorical Data Analysis - LECTURE 1
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
January 7, 2015
Vital statistics
My oce: Basic Science Building 255
Oce Hours:

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
Feburary 9, 2015
Poisson Loglinear Models for Count Data
For some categorical response vari

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
February 2, 2015
Components of a GLM
Generalized linear models (GLMs) : extension of ordina

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
March 16, 2015
Example
The following table is from a retrospective study of lung cancer
and

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
March 23, 2015
Using Models to Improve Inferential Power
Directed Alternatives:
Consider an

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
January 8, 2015
Vital statistics
My oce: Basic Science Building 255
Oce Hours: Fridays, 1:0

Maximum Likelihood Estimation
Let y denote the realized values of a set of observations and
P(Y, ) denote the joint density of Y = y and = (1 , . . . , q )
with belonging to some subset U Rq . The Likelihood of
given the observations is dened to be a fun

Discrete Sampling Distributions (CDA p.58)
The Bernoulli and Binomial distributions:
Bernoulli: Let Y denote a r.v. with support cfw_0, 1. Let y
denote the observed value of Y for a single
experiment. Suppose Pr(Y = 1) = and
Pr(Y = 0) = 1 . Then Y is call

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
Feburary 23, 2015
Categorical Predictors
Qualitative explanatory variables called factors
L

Welcome to BIST-512
Categorical Data Analysis
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
Feburary 9, 2015
Poisson Loglinear Models for Count Data
For some categorical response vari

Welcome to BIST-512
Categorical Data Analysis - LECTURE 2
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
January 12, 2015
Independence of Categorical Variables
When both X and Y are re

Welcome to BIST-512
Categorical Data Analysis - LEC9 : Loglinear models
Jaeil Ahn
Department of Biostatistics, Bioinformatics, and Biomathematics
Georgetown University
ja1030@georgetown.edu
March 30, 2015
1/58
Loglinear Models for 2-Way Tables
Consider an