metaSEM: Possible feature request

2 replies [Last post]
dstanley4's picture
Offline
Joined: 08/07/2012

Hi Everyone,
I keep enjoying using metaSEM. Right now, I'm using meta3 to conduct a meta-analysis where the studies are grouped by country. I would like to get a meta-analytic estimate of the mean for each country. In HLM, I would ask for the Level-3 residuals and combine that with the Intercept to get the meta-analytic mean for each country.

Is it possible to do this with metaSEM? If not, is there any chance this could be added as a feature in future updates!

Cheers,
David

Mike Cheung's picture
Offline
Joined: 10/08/2009
Thank you for the suggestion.

Thank you for the suggestion. After fitting the three-level model, the parameter estimates and their sampling variances can be obtained by:
model1 <- meta3(y=y, v=v, cluster=District, data=Cooper03)
coef(model1)
vcov(model1)

In principle, it is possible to compute the residuals or factor scores when the parameter estimates and their sampling variances are given. I have not checked the literature in details. The articles by Raudenbush and Bryk (1985) and Robinson (1991) semm to be relevant. I will put it into my to-do list. However, I cannot gurantee when I can do it. My top priorities are to formulate various meta-analytic models as structural equation models.

BTW, the main reference for meta3() is now Cheung (in press).

Cheung, M.W.L. (in press). Modeling dependent effect sizes with three-level meta-analyses: A structural equation modeling approach. Psychological Methods. Available at https://dl.dropboxusercontent.com/u/25182759/Modeling%20dependent%20effe...

Raudenbush, S. W., & Bryk, A. S. (1985). Empirical Bayes meta-analysis. Journal of Educational Statistics, 10, 75–98.

Robinson, G. K. (1991). That BLUP is a good thing: The estimation of random effects. Statistical Science, 6, 15–32.

dstanley4's picture
Offline
Joined: 08/07/2012
Thanks

Thanks for considering the suggestion - and the reading list.
I was particularly delighted to read:

Cheung, M.W.L. (in press). Modeling dependent effect sizes with three-level meta-analyses: A structural equation modeling approach. Psychological Methods.

The corresponding website you created is excellent!

Cheers,
David