3-Day Short-Course on Panel Data Methods for the Analysis of Health Data

18^th - 20^th July 2012

About the Presenters

Pravin K. Trivedi

has held Distinguished Professor rank at Indiana University and has taught econometrics at Australian, European and North American Universities. With Colin Cameron he has co-authored two leading graduate texts, Microeconometrics: Methods and Applications (2005) and Microeconometrics Using Stata (2009) and the Econometric Society Monograph, Regression Analysis of Count Data Second edition (2012). He is a past Co-Editor of Econometrics Journal and Associate Editor of Journal of Applied Econometrics. His main research interests are in microeconometrics and health economics.

Information about his academic activities and his CV can be downloaded from his homepage.

http://mypage.iu.edu/~trivedi

3-Day Short-Course on Panel Data Methods for the Analysis of Health Data

About the Workshop

This is a three-day intensive course on econometric methods for modeling individual and household level longitudinal (panel) data from health economics involving discrete choice, and both discrete and continuous outcome variables. While many issues in panel data analysis are generic, unique features of healthcare data have led to development of many econometric and statistical techniques within the sub-field of health econometrics. The course will cover both linear and nonlinear panel data models, with fixed and random effects and conditionally-correlated random effects. Extensions of basic models will cover endogenous regressors, separable and non-separable unobserved heterogeneity, dynamic effects, zero-inflation and over-dispersion. Emphasis will be on operational methods. For illustration in class and replication in practical data analysis sessions, real data sets from published empirical research will be used.

Course Outline

Course instruction will take the form of complementary lectures and tutorial sessions. The tutorial sessions will give participants an opportunity to apply methods introduced in class using some key data sets and Stata 12.0 package. The course coverage will include the following topics:

Basic Linear Panel Models

Some key features of health expenditure data; target parameters and relationships; some well-known data sets
Long vs. short panels; balanced vs. unbalanced panels; attrition bias
Fixed and random effect models; population averaged models; incidental parameters; identification of target parameters; transformation methods; OLS and GLS estimation; robust inference in presence of heteroskedasticity and serial correlation

Extensions of Linear Panel Models

Weak and strict exogeneity assumptions
Endogenous regressors and panel GMM estimation
Two-step GMM estimation
Dynamics and lagged variables (Arellano/Bond and related estimators)
Initial conditions problem and CCRE model
Mixed linear panel models; clustered data

Selected Nonlinear Panel Models

Binary and counted outcomes
Pooled models and mixture models with unobserved heterogeneity
Fixed and random effect binary outcome models
Conditional likelihood methods
Panel logit, tobit and Poisson regression
Nonlinear GMM estimation for exponential mean models
Difference transformations and conditional moment restrictions

Applications and Data Analysis

1.Stata’s xt commands for linear and nonlinear models
2.Panel data management and data transformations
3. xtreg, xtivreg commands in Stata
4.Specification tests for linear panels
5. Logit and Poisson regression models
6.Stata’s programmable GMM command
7. xtlogit and xtpoisson commands

Who Should Attend

The course is aimed at graduate students, researchers, economists, statisticians and consultants from private and public sector organisations, mainly in the health and allied services sector. Participants are expected to have an understanding of microeconomics and econometrics similar to that of an economics graduate of an Australian University.

On this site