Confidence intervals for the overall effect size in random-effects meta-analysis.
read more
Citations
Outlier and influence diagnostics for meta‐analysis
The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method
Methods to estimate the between-study variance and its uncertainty in meta-analysis
Conducting indirect-treatment-comparison and network-meta-analysis studies: report of the ISPOR Task Force on Indirect Treatment Comparisons Good Research Practices: part 2.
References
Statistical Power Analysis for the Behavioral Sciences
Meta-Analysis in Clinical Trials*
Statistical Methods for Meta-Analysis
Statistical Methods for Meta-Analysis.
Related Papers (5)
Frequently Asked Questions (13)
Q2. Why did the z distribution CI overstate the nominal confidence level?
The overstatement of the nominal confidence level obtained with the HM and SJ estimators was due to the fact that both of them are nonnegative estimators of 2 and, as a consequence, when 2 0, they are positively biased, leading to CIs that are too wide.
Q3. What is the expected performance of the weighted variance CI?
It is expected that the performance of the weighted variance CI obtained with the d index will be similar to that of meta-analyses that use other effect-size indices, provided they are relatively unbiased and follow an approximately normal distribution.
Q4. What is the main consequence of assuming a standard normal distribution to obtain a CI?
The main consequence of assuming a standard normal distribution to obtain a CI for ̂ with Equation 9 is that its actual coverage probability is smaller than the nominal confidence level, the width of the CI being too narrow.
Q5. What are the weights that can be obtained in real meta-analyses?
in real meta-analyses, the only weights that can be obtained are the estimated weights, ŵi, which have been calculated here for eight different heterogeneity variance estimators.
Q6. How can the authors estimate the optimal weights of the variance in a meta-analysis?
Once the authors have an unbiased sampling variance estimator, ̂i 2, to be applied in each study and a heterogeneity variance estimator, ̂2, the optimal weights, wi, can be estimated by ŵi 1/ ̂ 2 ̂i 2 .
Q7. What is the HE estimator used to test the homogeneity hypothesis?
Another estimator of the heterogeneity variance in metaanalysis, recently proposed by Sidik and Jonkman (2005), also yields nonnegative values.
Q8. Why did previous simulations not manipulate the parametric mean effect size?
previous simulation stud-ies have not manipulated the parametric mean effect size, , because it is expected that CIs calculated from z and t distributions should be invariant to a location shift (Brockwell & Gordon, 2001; Sidik & Jonkman, 2005).
Q9. What is the first iterative estimate of the heterogeneity variance?
The second iterative estimator of the heterogeneity variance in a random-effects model is based on restricted maximum likelihood estimation (REML).
Q10. What is the CI for the overall effect size in a random-effects meta-?
Heterogeneity Variance EstimatorsTo calculate a CI around the overall effect size in a meta-analysis where a random-effects model is assumed, an estimate of the heterogeneity variance is needed.
Q11. What is the advantage of assuming a t distribution for with k ?
Although the authors have focused on how to obtain a CI for the overall effect size, another advantage of assuming a t distribution for ̂ with k – 1 degrees of freedom and the weighted sampling variance, V̂w(̂), is that it is possible to test the null hypothesis of a parametric effect size equal to zero (H0: 0) with the test statistic T ̂/ V̂w ̂ .
Q12. What is the common method used to estimate the variance of a meta-analysis?
For most of the effectsize indices usually applied in meta-analysis, unbiased estimators of the sampling variance, ̂i2, have been derived, and several estimators can be found in the literature to estimate the heterogeneity variance in a meta-analysis, ̂2 (Sidik & Jonkman, 2007; Viechtbauer, 2005).
Q13. What was the sample size distribution used in their simulations?
The sample size distribution used in their simulations was obtained from a review of the meta-analyses published in 18 international psychological journals, with a Pearson skewness index of 1.464 (for more details, see Sánchez-Meca & Marı́nMartı́nez, 1998).