OK, would be a relevant action for the future while in other stat. programs support such models
Kind regards,
Gert
Gert Rijlaarsdam
Original Message:
Sent: 4/24/2023 3:34:00 PM
From: Rick Marcantonio
Subject: RE: compare regression models in meta analyses
The statistician's response:
"I agree that the model selection feature could be a future direction to enhance the Meta-regression procedure. But for now, we do not support in META REGRESSION."
------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------
Original Message:
Sent: Mon April 24, 2023 01:22 PM
From: Gert Rijlaarsdam
Subject: compare regression models in meta analyses
What you want todo is to see whether the addition of components improve the explanation of the variance in the effect sizes. it is not standard procedure, yet, but is an elegant way to create the best explanatory model, and to report the estimates based on that best model, I think?
------------------------------
Gert Rijlaarsdam
Original Message:
Sent: Mon April 24, 2023 11:22 AM
From: Rick Marcantonio
Subject: compare regression models in meta analyses
The statistician replies:
"The goal of meta-regression is not to compare different models but to see how much variation of effect sizes could be explained by a certain set of predictors. That's the reason why we did not add those goodness-of-fit statistics.
"If the customer is really interested in comparing models, I believe he or she may resort to the other procedures like REGRESSION or GLM by specifying the effect size estimates as a target or dependent variable."
------------------------------
Rick Marcantonio
Quality Assurance
IBM
Original Message:
Sent: Fri April 21, 2023 05:32 PM
From: Gert Rijlaarsdam
Subject: compare regression models in meta analyses
I try to run a series of regression models in meta-analysis, and compare the fit. Although Maximum Likelihood is one of the option for the estimator, I cannot find a goodness off fit index of the model as a whole, to compare with an other -nested – model. It is possible in R, so I assume it must be possible in spss?
Kind regards,
Gert