Unformatted text preview:

Slide 1AgendaMeta-RegressionFixed-Effect ModelFixed-Effect ModelFixed-Effect Model ANOVA TableRandom-Effects ModelRandom-Effects ModelRandom-Effects Model FitProportion of Covariate Explained Variance Variance Explained by CovariateToday’s First In-Class ActivityComplex Data StructuresIndependent Subgroups within a StudyCombining Across SubgroupsCombining Across SubgroupsMultiple Outcomes or Time-Points within a StudyCombining Outcomes or Time-PointsCombining Outcomes or Time-PointsCombining Outcomes or Time-PointsComparing Outcomes or Time-Points within a StudyComparing Outcomes or Time-PointsMultiple Comparisons within a StudyToday’s Second In-Class ActivityEVAL 6970: Meta-AnalysisMeta-Regression and Complex Data StructuresDr. Chris L. S. CorynSpring 2011Agenda•Meta-regression–In-class activity•Complex data structures–In-class activityMeta-Regression•Used to estimate the impact/influence of categorical and/or continuous covariates (moderators) on effect sizes or to predict effect sizes in studies with specific characteristics•A ratio of 10:1 (studies to covariates) is recommendedFixed-Effect ModelRegression of Latitude on Log risk ratioLatitudeLog risk ratio8.80 13.84 18.88 23.92 28.96 34.00 39.04 44.08 49.12 54.16 59.200.600.340.08-0.18-0.44-0.70-0.96-1.22-1.48-1.74-2.00Fixed-Effect ModelANOVA informationFixed-Effect Model ANOVA TableModel ()121.49992 1 0.00000Residual ()30.73309 11 0.00121Total () 152.23301 12 0.00000121.49992 1 0.0000030.73309 11 0.00121152.23301 12 0.00000• , means that the total variance is greater than would be expected based on within-study error• , means that the relationship between the covariate and the effect is greater than would be expected by chance• , means that even with the covariate in the model, some of the between-studies variance is unexplainedRandom-Effects ModelRegression of Latitude on Log risk ratioLatitudeLog risk ratio8.80 13.84 18.88 23.92 28.96 34.00 39.04 44.08 49.12 54.16 59.200.600.340.08-0.18-0.44-0.70-0.96-1.22-1.48-1.74-2.00Random-Effects ModelRandom-Effects Model Fit•Tests of the model–Simultaneous test that all coefficients (excluding intercept) are zero–Goodness of fit test that all unexplained variance is zero=CHIDIST(,)•?Proportion of Covariate Explained Variance•In meta-analysis, the total variance includes both variance within studies and between studies•Study-level covariates explain only the between-studies portion of the variance�2=1−(������������2������2)Calculating ?������2 Use the fixed-effect meta-analysis results(not meta-regression results)Calculating ?������������2 Results from random-effectsmeta-regression usingmethod of momentsVariance Explained by CovariateToday’s First In-Class Activity•From the “BCG Meta-Regression.CMA” data set–Using a risk ratio as the effect size, conduct a random-effects meta-regression (with method of moments) regressing latitude on the risk ratio–Write the regression equation, calculate the -test to estimate the impact of the slope, compute the LL and UL of , and calculate –Interpret and explain the results•?Complex Data Structures•Main categories of complex data structures–Independent subgroups within a study–Multiple outcomes or time-points within a study–Multiple comparisons within a study•The first two are (relatively) easily handled in Comprehensive Meta-Analysis 2.0Independent Subgroups within a Study•When two or more independent subgroups (each of which contribute unique information) are reported within the same study, the options are1. Compare effects between subgroups•For two subgroups, -test•For two or more subgroups, -test based on ANOVA•-test for heterogeneity 2. Compute a summary effect for all subgroups combined•?Combining Across Subgroups•Option 1a (effect size is computed within subgroups)–Treat each subgroup as a separate study•Interest is in between-subgroup variation•Option 1b (effect size is computed within studies)–Compute a composite score and use the composite score for each study as the unit of analysis•Interest is in between-study variationCombining Across Subgroups•Option 2 (ignore subgroup membership)–Collapse across subgroups to compute a summary effect size and variance–Subgroup membership is considered unimportant and is ignored (and its variance is not part of the summary effect size or standard error)–Essentially a main effect meta-analysisMultiple Outcomes or Time-Points within a Study•When a study reports data on more than one outcome, or over more than one time-point, where outcomes or time-points are based on the same participants (i.e., dependent), the options are1. Compute a composite effect size accounting for the correlation between outcomes or time-points2. Compute a difference between outcomes or time-points accounting for the correlation between outcomes or time-pointsCombining Outcomes or Time-Points•The effect size for two outcomes or time-points is computed as•With variance of the combined mean´� =12(�1=�2) ��1+��2+2�√��1√��2�´�=14¿Combining Outcomes or Time-Points•For more than two outcomes or time-points•With variance of ´� =1�(∑����) �´�=(1�)2(∑�=1���+∑�≠ �(���√��1√��2))Combining Outcomes or Time-Points•The problem is that often is not known (e.g., not reported in a study)•If is unknown, the only solution is to use a plausible value or range of values (sensitivity)–Similarity (or dissimilarity) of outcomes–Time elapsed between time-points and stability of relative scores over time•By default, Comprehensive Meta-Analysis 2.0 sets to 1.00 (which may overestimate the variance and underestimate precision)•?Comparing Outcomes or Time-Points within a Study•The effect size for the difference between two outcomes or time-points is computed as•With variance�����=�1−�2 �����=��1+��2− 2�√��1√��2Comparing Outcomes or Time-Points•As before, the problem is that often is not known (e.g., not reported in a study)•If is unknown, the only solution is to use a plausible value or range of values (sensitivity)•By default, Comprehensive Meta-Analysis 2.0 sets to 0.00 (which may overestimate the variance and underestimate precision of the difference)•?Multiple Comparisons within a Study•When a study


View Full Document

WMU EVAL 6970 - Lecture Notes

Download Lecture Notes
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Lecture Notes and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Lecture Notes 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?