A complete five-day course on the analysis of discrete data using SAS®
Here are a few of the things you'll learn in this course
What's wrong with ordinary linear regression when the dependent variable is a dichotomy.
Why ordinary regression is sometimes OK.
The easy and intuitive way to interpret logit coefficients.
Why logit coefficients are inherently standardized.
How to analyze contingency tables with a logistic regression program.
How to compare logit coefficients across groups, while adjusting for unobserved heterogeneity.
Why chi-square statistics should sometimes be ignored.
Why logistic regression is usually preferred to probit regression.
When to use the complementary log-log model.
Why logistic regression models sometimes fail to converge--and what you can do about it.
How sampling on the dependent variable can give sometimes give you better estimates.
How to estimate and interpret a multinomial logit model.
How you can fit a multinomial model with a binary logit program.
How to choose among three different approaches to ordinal dependent variables.
How to control for all constant characteristics of individuals using panel data.
Why marginal tables are so important in log-linear analysis.
How to convert a log-linear model to a logit model and vice versa.
Who should attend?
If you need to analyze categorical data and have a basic statistical background, this course is for you. You should have a good working knowledge of the principles and practice of multiple regression, as well as elementary statistical inference. But you do not need to know matrix algebra, calculus, or likelihood theory.
The course will stress generalized regression models with categorical dependent variables. Nevertheless, there will also be some treatment of two-way contingency tables in which neither variable has causal or predictive priority.
The course does emphasize SAS® rather heavily. No previous knowledge of SAS is assumed, however. Furthermore, nearly all the techniques taught in the course can be translated fairly easily to other packages.
Course Outline
1. Traditional methods for categorical data analysis 2. Review of linear model 3. OLS regression with categorical covariates 4. Dichotomous dependent variables in OLS regression 5. Heteroscedasticity and nonnormality 6. Weighted least squares estimates of linear probability model 7. Nonlinearity 8. The logit model 9. Estimating the logit model with grouped data 10. Estimating the logit model with ungrouped data 11. Maximum likelihood estimation 12. Interpreting logit coefficients 13. Similarities with multiple regression 14. Nonconvergence of ML estimates 15. Probit model and other link functions 16. Latent variable interpretation of the models 17. Unobserved heterogeneity 18. Logit and probit analysis for contingency tables 19. Goodness-of-fit tests for grouped data 20. Multinomial response models : unordered case 21. Logit models for ordered polytomies 22. Uniform association model 23. Cumulative logit and probit 24. Continuation ratio models 25. Latent variable interpreation 26. Response-based sampling 27. Poisson regression 28. Panel data 29. Random effects models 30. GEE estimation 31. Fixed effects logit model 32. Event history models 33. Discrete choice models 34. Loglinear models for contingency tables 35. Importance of marginal tables in loglinear models 36. Special loglinear models for mobility (and other 2-way) tables
Computing
The course will focus on four SAS® procedures for categorical data analysis: LOGISTIC, SURVEYLOGISTIC, GENMOD, and GLIMMIX. At least one hour each day is devoted to carefully structured and supervised assignments on personal computers. Additional time is available for exploring other sample data sets. Or you can bring your own data and try out new techniques as you learn them.
Location, format, materials
The course meets Monday through Friday at Temple University Center City, located in the heart of Philadelphia.
The textbook for the course is Professor Allison's book, Logistic Regression Using SAS® . In addition to the textbook, participants will receive a 100-page manual containing detailed lecture notes (with equations and graphics), examples of computer printout, and many other useful features. This book frees participants from the distracting task of notetaking.
Evaluations from Previous Participants
Participants in the July 2009 seminar were asked to rate the course on a scale of 1 (worst) to 10 (best). The average score for 33 respondents was 8.1. The following comments were also received:
“An excellent course. I've been back at work for several weeks now and keep discovering new ways that the course is helping me in my research.”
Louise Russell, Rutgers University
“Dr. Allison thoroughly knows the course material and is able to speak to the nuances of the topics. The hands-on experience in the labs also added to my experience.”
Dirk Matson
“I really enjoyed the course. It is ideally suited for those who use SAS and its statistical components on a regular basis and could use a more detailed explanation of the procedures. You will leave the course more confident in your skills.”
Anna Kaltenboeck
“This is very useful for any data analyst. Dr. Allison made this statistics course very interesting and explained things with practical examples. My apprehension about a 5-day statistics course was proved wrong, as this was very enjoyable.”
Suneel Guda, Schering-Plough
“I gained a deeper understanding of categorical data analysis, its theoretical development and its practical application. This course is excellent and highly recommended to anyone who wants to feel confident in categorical data analysis. Professor Allison offers a balanced approach to teaching both basic and cutting-edge knowledge.”
Antje Daub, Case Western Reserve University
“It is very informative and deepens my knowledge of categorical analysis.”
Yan Liu, Internal Revenue Service
“Dr. Allison is very knowledgeable about the subject area and provides numerous examples for modeling different kinds of data.”
