Contribution to the discussion of “When should meta‐analysis avoid making hidden normality assumptions?”

Pauly, Markus; Welz, Thilo

doi:10.1002/bimj.201800184

Cited by 2 publications

(3 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are several ways to estimate the covariance of β in ( 5)-( 6). Here we focus on six heteroscedasticity consistent (HC) estimators denoted by HC 0 , HC 1 , ..., HC 5 and the Knapp-Hartung (KH) estimator (Knapp & Hartung, 2003), which performed well in a meta-analytic context in previous research (Viechtbauer et al, 2015;Welz, 2018;Welz and Pauly, 2020;Welz et al, 2021). In the following section we introduce HC 0 , HC 1 , HC 2 according to MacKinnon and White (1985), HC 3 , HC 4 according to Cribari-Neto (2004) and HC 5 according to Cribari-Neto et al (2007) if not stated otherwise.…”

Section: Estimators For the Covariance Of βmentioning

confidence: 97%

“…Contributing to the discussion "When should meta-analysis avoid making hidden normality assumptions?" (Jackson and White, 2018; Pauly and Welz, 2018;Röver and Friede, 2018) we consider different distributions for the u i 's: The random effects are chosen as u i = √ τ 2 q i , where τ 2 ∈ {0.1, ..., 0.9} and the q i 's are independently sampled from either a standard normal-or a standardized exponential-, Laplace-, log-normal-or t 3 -distribution. Here t 3 denotes the t distribution with three degrees of freedom.…”

Section: Sample Sizesmentioning

confidence: 99%

See 1 more Smart Citation

Robust Confidence Intervals for Meta-Regression with Correlated Moderators

Knop¹,

Pauly²,

Friede³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Mixed-effects meta-regression models provide a powerful tool for evidence synthesis. In fact, modelling the study effect in terms of random effects and moderators not only allows to examine the impact of the moderators, but often leads to more accurate estimates of the involved parameters. Nevertheless, due to the often small number of studies on a specific research topic, interactions are often neglected in meta-regression. This was also the case in a recent meta-analysis in acute heart failure where a significant decline in death rate over calendar time was reported. However, we believe that an important interaction has been neglected. We therefore reanalyzed the data with a meta-regression model, including an interaction term of the median recruitment year and the average age of the patients. The model with interaction suggests different conclusions.This led to the new research questions (i) how moderator interactions influence inference in mixed-effects meta-regression models and (ii) whether some inference methods are more reliable than others. Focusing on confidence intervals for main and interaction parameters, we address these questions in an extensive simulation study. We thereby investigate coverage and length of seven different confidence intervals under varying conditions. We conclude with some practical recommendations.

show abstract

Section: Estimators For the Covariance Of βmentioning

confidence: 97%

Section: Sample Sizesmentioning

confidence: 99%

Robust Confidence Intervals for Meta-Regression with Correlated Moderators

Knop¹,

Pauly²,

Friede³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…This becomes especially important under adverse conditions, such as non-normally distributed effect sizes and/or unbalanced study sizes or arms. As already shown, 11 such circumstances can lead to poor control of type 1 error and/or poor coverage of confidence intervals when using standard meta-analytic techniques. For this paper, we therefore investigate the performance of the different estimators in different scenarios, utilizing both standardized mean differences and log-odds-ratios as effect measures.…”

mentioning

confidence: 99%

A simulation study to compare robust tests for linear mixed‐effects meta‐regression

Welz

Pauly

2020

Research Synthesis Methods

Self Cite

View full text Add to dashboard Cite

The explanation of heterogeneity when synthesizing different studies is an important issue in meta-analysis. Besides including a heterogeneity parameter in the statistical model, it is also important to understand possible causes of between-study heterogeneity. One possibility is to incorporate study-specific covariates in the model that account for between-study variability. This leads to linear mixed-effects meta-regression models. A number of alternative methods have been proposed to estimate the (co)variance of the estimated regression coefficients in these models, which subsequently drives differences in the results of statistical methods. To quantify this, we compare the performance of hypothesis tests for moderator effects based upon different heteroscedasticity consistent covariance matrix estimators and the (untruncated) Knapp-Hartung method in an extensive simulation study. In particular, we investigate type 1 error and power under varying conditions regarding the underlying distributions, heterogeneity, effect sizes, number of independent studies, and their sample sizes.Based upon these results, we give recommendations for suitable inference choices in different scenarios and highlight the danger of using tests regarding the study-specific moderators based on inappropriate covariance estimators.

show abstract

Contribution to the discussion of “When should meta‐analysis avoid making hidden normality assumptions?”

Cited by 2 publications

References 6 publications

Robust Confidence Intervals for Meta-Regression with Correlated Moderators

Robust Confidence Intervals for Meta-Regression with Correlated Moderators

A simulation study to compare robust tests for linear mixed‐effects meta‐regression

Contact Info

Product

Resources

About