Correlation is not causation, nor is conducting a correlational meta-analysis in MetaReviewer cause for concern! We introduce a correlational meta-analysis coding form template – complete with an updated ES section that smoothly handles correlational data and longitudinal study designs – to expand MetaReviewer’s capability of handling any type of review and meet your data needs.
Project Spotlight: Impulsivity and School Violence
Our new series highlights both the substantive content and unique data collection processes (and MetaReviewer updates) for several ongoing projects being conducted using MetaReviewer. First up, we spotlight what a review team at American Institutes for Research (AIR) is investigating: using longitudinal data to explore the association between impulsivity predicting future aggression. The project, “Investigating Impulsivity as a Root Cause of School Violence: A Systematic Review and Meta-Analysis,” is sponsored by the National Institute of Justice, and full text coding is currently underway. The project is led by one of MetaReviewer’s analysts, Laura Michaelson.
Why longitudinal? Researchers studying the relationship between impulsivity and school violence measure and report both cross-sectional and longitudinal correlations. By examining longitudinal associations between impulsivity measures (e.g., risk taking, emotional regulation) and aggression (e.g., bullying, delinquent offenses) in the same group of participants across the developmental span, the team aims to uncover if/the extent to which impulsivity in earlier developmental phases has robust predictive validity for aggressive behaviors in later childhood or teen years. The review team aims to influence how practitioners and policymakers interpret and use standard impulsivity measurements as potential risk identifiers.
A longitudinal correlational meta-analysis shares similarities with other types of systematic reviews and meta-analyses. Both review types, for example, capture demographic information of the sample and the nature of outcome constructs and measures. Rather than needing to identify and code key characteristics of an intervention or exposure to a condition and comparing outcomes between groups assigned or exposed to various conditions, however, a longitudinal correlational analysis pairs two “waves,” or data collection timepoints, in analyses that includes the same participants over time.
What's in a wave?
A wave, timepoint, snapshot…a rose by any other name. We’ve gone with “wave,” but whatever terminology suits your needs will do. When determining how to analyze longitudinal data, there are multiple timepoints at which data is collected. Depending on your analyses, the wave will correspond to various predictor and outcome variables. For the impulsivity meta-analysis, impulsivity measured at wave 1 (first data collection point) might be paired with bullying measured at wave 5 (five years after wave 1). With MetaReviewer’s “Add Column” feature, you can add as many waves as you need.
Varying variables, or, which came first?
For correlational analyses, defining predictor and outcome constructs and variables is key to not only answering your research questions, but in setting yourself up for success once you get to the ES page. We’ve created forms for both predictor and outcome variables, set up similarly to how waves are captured, for you to enter your specific data. At this stage, you don’t need to worry about defining when variables were measured (that’s what the waves page is for!), only whether you’ve identified them as predictors or outcomes.
Pulling it All Together
You’ve got your waves, predictors, and outcomes entered. Next up, ES calculation! The ES form in the new template functions almost identically to those in most existing MetaReviewer templates, but rather than identifying two groups, an outcome and a timepoint, for correlational analyses, you’ll enter your sample information, and two waves and one outcome and one predictor. This flexible format allows you to mix and match any combination of sample IDs, wave IDs, and predictor and outcome IDs.
We’ve also streamlined ES selection options, all with the same conditional follow-up entry items that you’ve come to love about MetaReviewer:
- OR: Posttest, Adjusted Logistic Regression Coefficient
- Corr: Bivariate Correlation
- Corr: Means and SD
- Corr: Frequencies, 2x2 Table
- Corr: Proportions, 2x2 Table
- Corr: Chi-square
- Corr: T Statistic (2 groups)
- Corr: T Statistic's p value (2 groups)
- Corr: Logged Odds Ratio
- Corr: Partial Correlation
- Corr: Semipartial Correlation
- Corr: Other - Team Calculated
Here's an example:
How Should Teams Use the Correlational Template?
In addition to already being fully capable of handling correlational meta-analyses, the new template can also be customized to meet your team’s needs for coding other within groups designs and longitudinal analyses. The new template option is as versatile and customizable as all other MetaReviewer templates, with all item options editable for your unique review design. We hope you’ll find this new template a useful addition to your project needs!
