This readme.txt file was generated on 20211005 by Robert J Buck ------------------- GENERAL INFORMATION ------------------- Ttle: Data and code for analyses in "The Use of Weighted Averages with Hedges: Is it Worth It?" Author: Robert J Buck, robert.j.buck@gmail.com Description: This repository contains R code to generate simulated meta-analyses and associated data R files (as text files) 1. DataDevelopmentForMEE.R Simulation results were run in blocks due to lengthy time to run them. This code describes parameter sets for blocks of simulations and runs those blocks. 2. librariesAndfunctionsForMEE.R Libraries used and functions created for simulations and anlyses. 3. TheoryAndSummaryDataForMEE.R Pulls together blocks of simulations, creates theoretical data; provides summary datasets. 4. FiguresForMEE.R Code for Figures and Table S1 presented in paper and supplementary materials. Data Files: 1. sigmaRatioData Theoretical data results, presumes tau2 and effect size are known. 2. modFnFCd.wtsKnown Simulation results for fixed effects, fixed N using grid method, weights known. 3. modFnRSd2.wtsKnown Simulation results for random effects, fixed N using grid method, weights known. 4. modFnRSd2.wtsUnKnown Simulation results for random effects, fixed N using grid method, weights unknown (Fig S8). 5. modFnFCd Simulation results for fixed effects, fixed N using grid method. 6. modFnRSd Simulation results for random effects, fixed N using grid method. 7. repData Simulation results replicating subset of modFnRSd, subsets modFnRSd2 and modFnRSd3 specifically, as columns. 8. modnbRSd Simulation results for random effects using negative binomial method. Variable List for Data Files: The data files modFnFCd.wtsKnwn, modFnRSd2.wtsKnwn, modFnFCd, modFnRSd, repData all have the same variables with the exception of files with FCd files do not have the RS variables described at the end of the list. Only a subset of variables is included in repData; these ar denoted by an *. 1. method* Weighted average method indicator 2. effectSize* Population Effect Size 3. popnICC* Population ICC = Popn between study/(between study + within study) 4. totalVar Total Variance = between study + within study variances. Set = 1 for all simulations. 5. M* Number of endpoints in a simulated meta-analysis 6. popnMn Popn Mean of simulated population 7. popnSD Popn SD of simulated population 8. Sn* Number of distinct sample sizes of endpoints in a simulated meta-analysis 9. N* Vector of distinct sample sizes of endpoints in a simulated meta-analysis 10. kN* Range of sample sizes = max(N)/min(N) 11. studySD True within study standard deviation 12. mn mean of simulated meta-analyses for given parameter settings 13. bias mn - popnMn 14. simSD standard deviation of simulated meta-analyses for given parameter settings 15. sumWtsSE.mn average of classic stdErr estimate of weighted average sqrt(1/sum(weights)) across simulations 16. sumWtsSE.sd StdDev of classic stdErr estimate of weighted average sqrt(1/sum(weights)) across simulations 17. hkSE.mn average of hauptmann-knapp stdErr estimate across simulations 18. hkSE.sd StdDev of hauptmann-knapp stdErr estimate across simulations 19. modhkSE.mn average of modified hauptmann-knapp StdErr estimate across simulations 20. du.eqWtsSE.mn Theoretical StdErr for unadjusted d with equal weights are known 21. du.exDhWtsSE.mn Theoretical StdErr for unadjusted d with exact known weights 22. dh.eqWtsSE.mn Theoretical StdErr for adjusted d with equal weights are known 23. dh.exDhWtsSE.mn Theoretical StdErr for adjusted d with exact known weights 24. rmse root mean squared error of simuklated meta-analyses for given parameter settings 25. simEffSize.mn average of estimated effect size using estimated mean and std dev using hauptmann-knapp 26. simEffSize.sd stdDev of estimated effect size using estimated mean and std dev using hauptmann-knapp 27. pctSig Percent of simulated meta-anlyses that were significant using H-K estimate for stdDev , i.e., had p- value < 0.05. 28. modPctSig Percent of simulated meta-anlyses that were significant using modified H-K estimate for stdDev, i.e., had p- value < 0.05. Random Effects only 29. reSE.mn estimate of SE across simulations = Square root of Mean of Variance estimates 30. remCIpctSig percent of simulations that were outside estimated 95% confidence interval. 31. tau2.mn mean of tau2 estimates from simulated meta-analyses for given parameter setting 32. tau2.sd Standard deviation of tau2 estimates from simulated meta-analyses for given parameter setting 33. pctTau2eq0 Percent of estimated tau2 = 0 in simulated meta-analyses for given parameter settings Generate HK-like estimate of variance, i.e., weighted variance calculation, using study variance only, no tau2. Not discussed in paper or supplementary materials. 34. eWthkSEobs.mn mean of HK-like estimate of variance 35. eWthkSEobs.sd standard deviation of HK-like estimate of variance repData only 36. metric Metric the value is measuring, e.g., bias, pctSig, simSD. 37. value Value from full data set. 38. value.rep replicate value. Variables for SigmaRatioData:Theoretical Results (tau2, effectSize known) 1. tau2 between study variance 2. effectSize Effect size for endpoint 3. n Sample size of endpoint study 4. varEX Exact Calculation for Variance Var(d) 5. varLS Large Sample estimate of Var(d) 6. varNE Hedges 1985 estimator of Var(d) where only use a(n) 7. Je Hedges adjustment parameter for effect size, exact formula 8. dlb lower bound of 95% Confidence Interval for d. 9. dub upper bound of 95% Confidence Interval for d. 10. varEXlb exact variance using dlb rather than effectSize in Var(d) calculation 11. varEXub exact variance using dub rather than effectSize in Var(d) calculation 12. wlb weight using tau2 and dlb rather than effectSize in Var(d) calculation 13. wub weight using tau2 and dub rather than effectSize in Var(d) calculation 14. aN a(N) from var(d) = a(N) + b(N)*effectSize^2 15. bNd2 b(N)*effectSize^2 16. w weight for exact variance of dH 17. ICC Intraclass correlation coefficient 18. ICClb theoretical lower bound for ICC 19. ICCub theoretical upper bound for ICC 20. invJe see code for remaining definitions. None are used for any results discussed in paper or supplementary materials. 21. JeRatio 22. wtJratio 23. Je2Ratio 24. wtJ2ratio ------------------- Version Notes ------------------- On 2022-01-19, "FiguresForMEE.r" was replaced with an updated version. The original version can be accessed at: https://hdl.handle.net/11299/224437.1