2
$\begingroup$

I'm planning to write a meta-analysis on a topic in which there are both clinical trials and structural equation modeling papers.

For clinical trials i can compute easily the effect size and analyze them with a fixed or random effects model, is there a way to calculate an effect size in a systematic way from a structural equation modeling result? Can i compute the effect size using the r of correlation in the models? Are there any reference regarding this approach?

I know that there are meta-analytic structural equation modeling approach but i want to conduct a "regular" meta-analysis with data from equation modeling studies

$\endgroup$
  • 1
    $\begingroup$ It will be useful to explain what a "regular" meta-analysis is. Do you want to summarize the findings in a structural equation model with one single effect size? $\endgroup$ – Mike Cheung Apr 20 '16 at 12:39
  • $\begingroup$ a regular meta-analysis is a meta analysis conducted with the canonic fixed or random effects model, not using a structural equation modeling approach. I'd like to extract an effect size regarding a specific relationship in the structural equation between a single variable in the equation and the output variable. $\endgroup$ – GGA Apr 20 '16 at 14:33
  • 1
    $\begingroup$ I'd have thought that for a meta-analysis you need an effect and a standard error. You can get them from a clinical trial, and you should be able to get them from an SEM. $\endgroup$ – Jeremy Miles Apr 21 '16 at 15:45
1
$\begingroup$

It depends on what metric of effect size you are using. If it's merely correlation from regression models versus SEM, remember that covariance-based SEM is similarly based in assumptions of linearity and normality, so standardized coefficients to get a correlation coefficient should compare favorably to standardized regression coefficients from linear regressions.

You may, however, want to use model-source as a covariate, just to check to see if you get different results from studies using SEMs versus linear regressions.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.