Meta-Analysis Resources

Tools for Those Who Summarize the Evidence Base

Resources and networking for those who conduct or interpret meta-analyses related to any phenomenon that is gauged in multiple studies.

Hi Dr. Cheung and MASEM users,

Our team is currently working to conduct a MASEM analysis on 40 studies assessing a multiple mediator model with treatment condition as a dichotomous predictor, posttreatment symptoms as the outcome, multiple mediators at posttreatment, and pretreatment symptoms and mediators as covariates. Some studies only reported data for some mediators, and other did not report data on any mediator, but all reported data for outcomes. We know that too much missing data can lead to not positive definite matrix.

Thus, we are coding all possible mediators now, but intend to either drop or combine some mediator categories that have too little data. We would appreciate guidance on how to assess how much data is enough within and between correlation matrices in order to run our analyses, so we can decide what mediator categories to drop or combine as well as determine which studies out of the 40 do not provide enough information to include in our analyses and should not be coded.

  • For example, at minimum should each cell of the pooled correlation matrix contain at least X% or another  percentage of the 40 studies?
  • Should each study provide data in at least Y% of the cells in the correlation matrix?

We would be very grateful for any estimate as to X and Y. Answers to these questions will help us to streamline our coding process.

In case this is helpful, we have attached my syntax and data file with sample sizes for 3 studies (test sample).

Thank you,

Mei Yi & Katherine

Views: 50


© 2018   Created by Blair T. Johnson.   Powered by

Badges  |  Report an Issue  |  Terms of Service