Goodness-of-fit test for meta-analysis

Abstract

Meta-analysis is a very useful tool to combine information from different sources. Fixed effect and random effect models are widely used in meta-analysis. Despite their popularity, they may give us misleading results if the models don’t fit the data but are blindly used. Therefore, like any statistical analysis, checking the model fitting is an important step. However, in practice, the goodness-of-fit in meta-analysis is rarely discussed. In this paper, we propose some tests to check the goodness-of-fit for the fixed and random effect models with assumption of normal distributions in meta-analysis. Through simulation study, we show that the proposed tests control type I error rate very well. To demonstrate the usefulness of the proposed tests, we also apply them to some real data sets. Our study shows that the proposed tests are useful tools in checking the goodness-of-fit of the normal models used in meta-analysis.

Interleukin 1 Polymorphisms Contribute to Intervertebral Disc Degeneration Risk: A Meta-Analysis

“…As goodness-of is fit for the fixed and random effect models with assumption of normal distributions in meta-analysis. Goodness-of-fit test was applied to check the model adequacy [ 29 ], and there was strong evidence that (all studies/data) fit the model very well.…”

“…R software was used to run Goodness-of-fit test to analyze data. Goodness-of-fit test was used to check how the models fit the data in meta-analysis [ 29 ], where Shapiro-Wilk test was applied in R. p_sw is the p-value that obtained by Shapiro-Wilk test with B = 100000 bootstraps. All analyses were performed using Review Manager 4.2 (Cochrane Collaboration, Oxford, UK) and STATA 12 (Stata, CollegeStation, TX) and R version 3.2.3(UOA, Nz).…”

Confirmatory Factor Analysis of Persian Version of Depression, Anxiety and Stress (DASS-42): Non-Clinical Sample

“…The goodness-of-fit evaluated using the following statistics: goodness-of-fit index (GFI > 0.85), adjusted goodness-of-fit index (AGFI > 0.80), nonnormal fit index (NNFI > 0.90), comparative fit index (CFI > 0.90), root mean square residual (RMSR < 0.10), normal chi-square (3 > χ 2 / df < 2) and root mean square error of approximation (RMSEA). Itis 90% interval confidence (31)(32)(33). The concurrent validity is investigated by the correlations between DASS-IR scores with BDI, MMPI-II (Pt, D, HS), and CAS.…”

Statistical power and utility of meta-analysis methods for cross-phenotype genome-wide association studies

“…The method is however applicable when all variants share the same genetic correlation across all traits (i.e., no subset specific effect is assumed), which is violated in most circumstances of cross-phenotype studies. In addition, misleading results can arise when using inadequately fitted meta-analysis models, thus it is recommended to perform the goodness-of-fit test before conducting meta-analysis [ 41 ]. Finally, integration of external genomic information in cross-phenotype meta-analysis is a largely unexplored territory.…”