Far East Journal of Theoretical Statistics
Volume 13, Issue 1, Pages 101 - 126
(May 2004)
|
|
IMPROVED TESTS OF HOMOGENEITY IN RANDOMIZED CONTROLLED MULTI-CENTER TRIALS WITH BINARY OUTCOME
Joachim Hartung (Germany) and Guido Knapp (Germany)
|
Abstract: In this
paper we present new tests of homogeneity for
the outcome measures risk difference and log
odds ratio in randomized controlled
multi-center trials provided that the
between-center variance is the parameter of
heterogeneity. Moreover, we show that the
commonly used test of homogeneity can be
improved in case the inverses of the variances
of the estimated parameters in each center are
estimated more efficaciously. But the new
proposed general c2
-test
has better properties concerning size and
power than the modified commonly used test
based on results of a simulation study.
Furthermore, one- and
two-sided confidence intervals for the
between-center variance are deduced. A
simulation study shows that these confidence
intervals satisfactorily keep the prescribed
confidence coefficient. |
Keywords and phrases: risk difference, odds ratio, test of homogeneity, random effects model, confidence interval for the between-center variance, meta-analysis. |
|
Number of Downloads: 277 | Number of Views: 664 |
|