by A. Kazemnejad, M. Mohebbi and M. Sanagoo.

Abstract: >Consider a test-retest study in which subjects are selected to participate in an intervention because one baseline measurement of a variable is within a particular range. For example screening specimens from extreme of a population. This requires taking measurements on selected subjects, usually without a parallel control group and, hence, it is necessary to adjust the effect for regression to the mean. We propose a two-step procedure to estimate an additive regression to the mean model. First estimate the model parameters by least square or moments. Second use these parameter estimates as initial values fit into maximum likelihood procedure to handling several kinds of sampling procedures: truncated, censored, selected and complete sampling. This method also does not have the restriction of knowing the exact value of the population mean or proportion of truncated population. Tests based on this new approach are also discussed. An example from

Key Words: Bivariate normal, Regression to the mean, Wald's test, Likelihood ratio test

A. Kazemnejad,
M. Mohebbi,
M. Sanagoo

Editor: Simos GMeintanis G,

READING THE ARTICLE: You can read the article in portable document (.pdf) format (145540 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2757 times since July 24, 2006.

Return to the InterStat Home Page.