Seminars

General Departmental Seminar Series

Joint Statistics/Biostatistics Seminar

Smooth Quantile Ratio Eestimation (SQUARE); Method and
Application to Costs of Smoking-Caused Diseases

Scott Zeger, PhD,
Dept. of Biostatistics, Bloomberg School of Public Health,
The Johns Hopkins University

October 6, 2004, 4 - 5 pm in 140 Bardeen Medical Laboratory, 1300 University Ave.

ABSTRACT

In comparing the mean health care costs for persons with major smoking-caused diseases (MSCD) to others without such diseases, the statistical problem is how to best estimate the difference in means for two right-skewed distributions from a relatively small sample of MSCD cases and a large sample of controls. Economists have historically favored log-normal models for such a problem but are more recently turning to semi-parametric methods. In this talk, we will present smooth quantile ratio estimation or SQUARE, an alternative that is more flexible and more efficient than parametric models in many situations. Statistical methods and theory will be summarized. SQUARE will be illustrated by an analysis of data from the National Medical Expenditure Survey (NMES). Extensions to regression will be briefly illustrated.

This seminar is jointly sponsored with the Department of Statistics.

Return to seminar list