AMS 573, Design and Analysis of Categorical Data
Measuring the strength of association between pairs of categorical variables. Methods
for evaluating classification procedures and inter-rater agreement. Analysis of the
associations among three or more categorical variables using log linear models. Logistic
regression.
Pre-requisite: AMS 572
3 credits, ABCF grading
Texts:
Categorical Data Analysis, by Alan Agresti, 3rd edition, Wiley Publisher, 2013; ISBN: 978-0-470-46363-5 (required)
Categorical Data Analysis Using SAS by Maura E. Stokes, Charles S. Davis, Gary G. Koch, 3rd edition, SAS publisher, 2012; ISBN: 978-1-60764-664-8 (optional/recommended)
Spring Semester
Learning Outcomes:
1) Demonstrate skills of working with various categorical data, including binary,
nominal, ordinal and count data:
* Expectation, variance, covariance and probability density function;
* Point estimation with maximal likelihood method;
* Hypothesis testing with Wald, score and likelihood ratio tests;
* Constructing confidence intervals based on Wald, score and likelihood ratio
test statistics.
2) Demonstrate skills with statistical inference for contingency tables (joint distribution
of categorical variables):
* Difference of proportions, relative risk and odds ratio;
* Chi-squared tests;
* Fisher’s exact test;
* McNemar test for matched pairs.
3) Demonstrate skills with statistical modeling for binary/nominal/ordinal response:
* Build and apply logistic regression, baseline category and cumulative logit
models;
* Maximal likelihood fitting and goodness of fit tests;
* Model diagnostic and model selection;
* Other link functions: log-log, complementary log-log.
4) Demonstrate skills with statistical modeling for count data:
* Build and apply log-linear models;
* Connection between log-linear and logit models;
* Model fitting and goodness of fit tests;
* Association graphs and collapsibility.
5) Demonstrate skills with proficient usage of standard statistical software tools
for categorical data analysis:
* Understanding of the assumptions, derivation and interpretation of results
from statistical analysis;
* Proficient in SAS procedures: FREQ, GENMOD, GLM and LOGISTIC.