[R-meta] Mean and SD or unstandardized regression coefficients?

[Lukasz Stasielowicz]

Dear Stefanou,

as Michael Dewey has pointed out, comparing regression coefficients 
based on models with different sets of variables can be difficult as the 
meaning of the coefficients changes. Furthermore, adding certain 
variables can introduce suppression effects, which can sometimes change 
the sign of the regression coefficient. In the meta-analytic context it 
could introduce spurious heterogeneity (e.g., negative and positive 
regression coefficients).

In addition, when one really wants to synthesize unstandarized 
regression coefficients then one has to ensure that the same instruments 
were used across studies or that the coding of the variables is 
consistent across studies, e.g. group is always coded as 0 = Control 
group and 1 = Intervention group, time scale is always the same (e.g. 
Study 1 - days, Study 2 - days vs. Study 1 - days, Study 2 - weeks) etc.


See also:
Becker, B. J., & Wu, M.-J. (2007). The synthesis of regression slopes in 
meta-analysis. Statistical Science, 22(3), 414–429. 
https://doi.org/10.1214/07-STS243


Thus, synthesizing means or mean differences seems to be preferable. If 
there are enough studies then one could conduct additional analyses, 
e.g. synthesizing regression coefficients based on regression models 
with the same set of variables.


Best wishes,
Lukasz
-- 
Lukasz Stasielowicz
Osnabrück University
Institute for Psychology
Research methods, psychological assessment, and evaluation
Seminarstraße 20
49074 Osnabrück (Germany)

Am 23.11.2021 um 12:18 schrieb r-sig-meta-analysis-request using r-project.org:
> Send R-sig-meta-analysis mailing list submissions to
> 	r-sig-meta-analysis using r-project.org
> 
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
> or, via email, send a message with subject or body 'help' to
> 	r-sig-meta-analysis-request using r-project.org
> 
> You can reach the person managing the list at
> 	r-sig-meta-analysis-owner using r-project.org
> 
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of R-sig-meta-analysis digest..."
> 
> 
> Today's Topics:
> 
>     1. Re: Mean and SD or unstandardized regression coefficients?
>        (Michael Dewey)
> 
> ----------------------------------------------------------------------
> 
> Message: 1
> Date: Mon, 22 Nov 2021 11:07:09 +0000
> From: Michael Dewey <lists using dewey.myzen.co.uk>
> To: Stefanou Revesz <stefanourevesz using gmail.com>, R meta
> 	<r-sig-meta-analysis using r-project.org>
> Subject: Re: [R-meta] Mean and SD or unstandardized regression
> 	coefficients?
> Message-ID: <45c857de-e284-0624-afb7-fe077fa5428c using dewey.myzen.co.uk>
> Content-Type: text/plain; charset="utf-8"; Format="flowed"
> 
> Dear Stefanou
> 
> I suspect there may be different opinions here but it seems to me that
> the advantage of the regression coefficients is that they provide a
> better estimate of the effect but their disadvantage is that if primary
> studies adjusted for different variables they are not estimating the
> same thing. So depending on what they adjusted for I might choose either
> of them under different circumstances.
> 
> Michael
> 
> On 21/11/2021 16:21, Stefanou Revesz wrote:
>> Dear Meta-Analysis Experts,
>>
>> My sample of longitudinal studies both provide the raw descriptive
>> statistics (mean, sd) for groups at different time points as well as
>> the unstandardized regression coefficients for group*time interactions
>> often adjusted for various confounding variables which may not be the
>> same across the studies.
>>
>> To meta-analyze these studies, is it better to meta-analyze the
>> adjusted unstandardized regression coefficients for group*time
>> interactions or the raw (i.e., unadjusted) descriptive statistics
>> (mean, sd) for groups at different time points?
>>
>> Thanks,
>> Stefanou
>>
>> _______________________________________________
>> R-sig-meta-analysis mailing list
>> R-sig-meta-analysis using r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>
>