Sankhya: The Indian Journal of Statistics

2001, Volume 63, Series B, Pt. 3, pp. 298--310

TESTING FOR HOMOGENEITY IN COMBINING OF TWO-ARMED TRIALS WITH NORMALLY DISTRIBUTED RESPONSES

By

JOACHIM HARTUNG

and

DOGAN ARGAÇ, University of Dortmund, Germany

SUMMARY. We deal with testing for homogeneity of mean differences in a series of independent two-armed trials with normally distributed outcomes assuming heteroscedastic variances both among the two arms in each trial and between the various trials. We consider tests for investigating this hypothesis under these nonstandard circumstances using classical analysis of variance techniques and concavity arguments. By way of simulation, the performance of the tests with respect to the attained significance levels and power is compared. The proposed tests keep their level and attain reasonable power.

AMS (1991) subject classification. Primary 62J10; secondary 62P10.

Key words and phrases. Heteroscedastic variances, homogeneity, meta-analysis, two-armed experiments, refined Cochran test, refined Welch test, Behrens-Fisher problem.

Full paper (PDF)