Longitudinal Data Analysis Using Stata

In this two day seminar, taught by Professor Paul Allison, you will learn how to do regression analysis of panel data—the most common type of longitudinal data—consisting of measurements of predictor and response variables at two or more points in time for many individuals. Panel data have two major attractions: the ability to control for unobservables, and the determination of causal ordering. But the downside is that repeated measurements typically violate assumptions of independence.

The course covers four methods for solving the problem of dependent observations: robust standard errors, generalized estimating equations, random effects (mixed) models and fixed effects models. You'll learn how to use these methods for quantitative outcomes, categorical outcomes, and count data outcomes.

Here are a few of the topics you won't want to miss:

• How to use panel data to control for unobserved variables.
• Why fixed effects methods often give very different results from random effects methods.
• How to reshape data from long form to wide form and back again.
• Why the default correlation structure for GEE is usually not the best.
• The difference between maximum likelihood and restricted maximum likelihood.
• How to estimate and interpret random coefficients models.
• Why first-order autoregressive structures are usually unsatisfactory.
• The difference between subject-specific coefficients and population-averaged coefficients, and why it matters.
• How to do longitudinal analysis using ordered logit or multinomial logit.

The course covers the following commands: reg, areg, xtreg, xtmixed, logit, ologit, clogit, mlogit, xtlogit, xtmelogit, xtgee, xtpoisson, xtmepoisson, xtnbreg, and reshape. Lecture notes using SAS are also available to registered participants.