Use of baseline data for estimation of treatment effects in the presence of regression to the mean

Shande Chen, C. Cox

Research output: Contribution to journalArticlepeer-review

21 Scopus citations

Abstract

We consider a regression to the mean problem with a very large sample for the first measurement and relatively small subsample for the second measurement, selected on the basis of the initial measurement. This is a situation that often occurs in screening trials. We propose to estimate the unselected population mean and variance from the first measurement in the larger sample. Using these estimates, the correlation between the two measurements, as well as an effect of treatment, can be estimated in simple and explicit form. Under the condition that the size of the subsample is of a smaller order, the new estimators for all the four parameters are as asymptotically efficient as the usual maximum likelihood estimators. Tests based on this new approach are also discussed. An illustration from a cholesterol screening study is included.

Original languageEnglish
Pages (from-to)593-598
Number of pages6
JournalBiometrics
Volume48
Issue number2
DOIs
StatePublished - 8 Jul 1992

Fingerprint

Dive into the research topics of 'Use of baseline data for estimation of treatment effects in the presence of regression to the mean'. Together they form a unique fingerprint.

Cite this