[R-meta] Question about MASEM with categorical predictors

Mike Cheung m|kew|cheung @end|ng |rom gm@||@com
Wed Nov 23 03:42:48 CET 2022


Dear Catia,

If I understand it correctly, you use hetcor() or other functions to
estimate a "correlation matrix" of continuous and binary variables and use
this correlation matrix in tssem1() and then tssem2().

The part about continuous variables should be fine. But I am not sure about
the part on binary and continuous variables. Moreover, the meaning of
including a latent variable with a threshold on gender is questionable.

The most defensible approach is (1) fitting the regression model in each
study, (2) estimating the path coefficients and their sampling covariance
matrix, and (3) conducting a multivariate meta-analysis on the path
coefficients.

Best,
Mike


On Tue, Nov 22, 2022 at 10:25 PM Catia Oliveira <catia.oliveira using york.ac.uk>
wrote:

> Just to make sure I understand, I would start by getting the correlations
> per study using the hetcor() (maybe the polychoric() from the psych package
> would be better since it computes a polychoric correlation matrix). I would
> then use  tssem1() to pool the correlation matrices across studies and then
> fit the model (Outcome ~ categorical variable + numerical variable +
> numerical variable) using sem(). Is that what you mean? I am sorry for
> taking so much out of your time but I have not been able to find any
> resources about this.
>
> Thank you,
>
> Catia
>
> On Tue, 22 Nov 2022 at 02:32, Mike Cheung <mikewlcheung using gmail.com> wrote:
>
>> Dear Catia,
>>
>> You may use the wls() function to conduct the stage two analysis in the
>> metaSEM package.
>>
>> However, It appears that hetcor does not compute the asymptotic sampling
>> covariance matrix of the estimated correlations. Thus, the statistical
>> inferences may be incorrect.
>>
>> If you want to use this approach, it is better to use the sem() function
>> in the lavaan package. I believe that it can generate the asymptotic
>> sampling covariance matrix of the estimated correlations.
>>
>> Best,
>> Mike
>>
>> On Mon, Nov 21, 2022 at 11:40 PM Catia Oliveira <
>> catia.oliveira using york.ac.uk> wrote:
>>
>>> Dear professor Mike,
>>>
>>> Thank you for your response. Wouldn't it be possible to run the path
>>> model I described if I used the hetcor() function in R on the raw data per
>>> study, where I would gather all the correlations between variables,
>>> including between gender and the numeric variables? Wouldn't this allow me
>>> to use the two staged approach you describe in "metaSEM: an R package for
>>> meta-analysis using structural equation modeling"?
>>>
>>> Thank you!
>>>
>>> Best wishes,
>>>
>>> Catia
>>>
>>>
>>>
>>> On Sun, 13 Nov 2022 at 04:02, Mike Cheung <mikewlcheung using gmail.com>
>>> wrote:
>>>
>>>> Dear Catia,
>>>>
>>>> If you have the raw data, you may use either a multiple-group SEM or
>>>> multilevel SEM (assuming all variables are in comparable scales across
>>>> studies). However, you may need to standardize or harmonize the variables
>>>> before the analyses if the variables are not directly comparable across
>>>> studies.
>>>>
>>>> It may be tricky to pool correlation matrices when there are
>>>> categorical variables.
>>>>
>>>> An alternative is to fit the regression model in each group and
>>>> meta-analyze the regression coefficients. There was some discussion in the
>>>> following paper.
>>>>
>>>> Cheung, M. W.-L., & Cheung, S. F. (2016). Random-effects models for
>>>> meta-analytic structural equation modeling: Review, issues, and
>>>> illustrations. Research Synthesis Methods, 7(2), 140–155.
>>>> https://doi.org/10.1002/jrsm.1166
>>>>
>>>> I hope it helps.
>>>>
>>>> Best,
>>>> Mike
>>>>
>>>> On Sun, Nov 13, 2022 at 11:00 AM Catia Oliveira <
>>>> catia.oliveira using york.ac.uk> wrote:
>>>>
>>>>> Thank you both for replying to my question.
>>>>>
>>>>> @Mike Cheung <mikewlcheung using nus.edu.sg> The categorical variable
>>>>> represents sex, so only female and male. We aim to include only studies
>>>>> that have reported all the predictors of interest in the meta-analysis, so
>>>>> we are not anticipating missing data. It is unclear whether we will have
>>>>> access to the raw data for all studies, as we will not be able to know
>>>>> until we start going through the literature, but first, we need to
>>>>> preregister the meta-analysis. However, we are trying to anticipate all
>>>>> situations. I have read about the work you have done with Susanne Jak, but
>>>>> it is not clear whether I could include categorical variables using that
>>>>> approach. It also seems to require a lot of data, which may not be our case.
>>>>> If we have access to the raw data and are able to fit the model on
>>>>> each dataset, do you have any suggestions for how to better analyse it?
>>>>> From what I've read it seems that we could analyse each factor loading on
>>>>> its own and run a meta-regression, would that be reasonable?
>>>>>
>>>>> Thank you.
>>>>>
>>>>> Best wishes,
>>>>>
>>>>> Catia
>>>>>
>>>>> On Sun, 13 Nov 2022 at 01:26, Mike Cheung <mikewlcheung using gmail.com>
>>>>> wrote:
>>>>>
>>>>>> Dear Catia,
>>>>>>
>>>>>> Could you be more specific about how the data look like? For example,
>>>>>> do you have the raw data? If not, what types of summary statistics do
>>>>>> you
>>>>>> have?
>>>>>>
>>>>>> How many levels are in V1? How many groups are? Are there incomplete
>>>>>> data?
>>>>>>
>>>>>> --
>>>>>> ---------------------------------------------------------------------
>>>>>>  Mike W.L. Cheung               Phone: (65) 6516-3702
>>>>>>  Department of Psychology       Fax:   (65) 6773-1843
>>>>>>  National University of Singapore
>>>>>>  http://mikewlcheung.github.io/
>>>>>> <http://courses.nus.edu.sg/course/psycwlm/internet/>
>>>>>> ---------------------------------------------------------------------
>>>>>>
>>>>>> On Sat, Nov 12, 2022 at 9:17 PM Lukasz Stasielowicz <
>>>>>> lukasz.stasielowicz using uni-osnabrueck.de> wrote:
>>>>>>
>>>>>> > Dear Catia,
>>>>>> >
>>>>>> > Disclaimer: I am not up-to-date with MASEM advances so perhaps
>>>>>> there is
>>>>>> > a more user-friendly solution. Consider checking recent work by Mike
>>>>>> > Cheung and Suzanne Jak for references about cutting-edge methods.
>>>>>> >
>>>>>> > If you have access to raw data then one could use the
>>>>>> parameter-based
>>>>>> > MASEM approach. Since it is possible to use categorical predictors
>>>>>> in
>>>>>> > lavaan/blavaan, one could fit the model separately for each sample
>>>>>> and
>>>>>> > then pool the estimates. This approach has a clear practical
>>>>>> limitation:
>>>>>> > all variables/categories need to be assessed in all studies, which
>>>>>> is
>>>>>> > not always the case.
>>>>>> >
>>>>>> > Alternatively, one could fit separate models for each category
>>>>>> (e.g.,
>>>>>> > women, men). After all, stratification is just another kind of
>>>>>> adjusting
>>>>>> > for variables.
>>>>>> >
>>>>>> > Some references about the parameter-based approach can be found in
>>>>>> this
>>>>>> > article:
>>>>>> > Cheung, M. W.-L. (2021). Meta-analytic structural equation modeling.
>>>>>> > Oxford Research Encyclopedia of Business and Management, Oxford
>>>>>> > University Press.
>>>>>> https://doi.org/10.1093/acrefore/9780190224851.013.225
>>>>>> >
>>>>>> >
>>>>>> > Best,
>>>>>> > Lukasz
>>>>>> > --
>>>>>> > Lukasz Stasielowicz
>>>>>> > Osnabrück University
>>>>>> > Institute for Psychology
>>>>>> > Research methods, psychological assessment, and evaluation
>>>>>> > Lise-Meitner-Straße 3
>>>>>> > 49076 Osnabrück (Germany)
>>>>>> > Twitter: https://twitter.com/l_stasielowicz
>>>>>> >
>>>>>> > On 06.11.2022 12:00, r-sig-meta-analysis-request using r-project.org
>>>>>> wrote:
>>>>>> > > 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. Question about MASEM with categorical predictors (Catia
>>>>>> Oliveira)
>>>>>> > >
>>>>>> > >
>>>>>> ----------------------------------------------------------------------
>>>>>> > >
>>>>>> > > Message: 1
>>>>>> > > Date: Sat, 5 Nov 2022 23:21:10 +0000
>>>>>> > > From: Catia Oliveira <catia.oliveira using york.ac.uk>
>>>>>> > > To: R meta <r-sig-meta-analysis using r-project.org>
>>>>>> > > Subject: [R-meta] Question about MASEM with categorical predictors
>>>>>> > > Message-ID:
>>>>>> > >       <CACw+TfdnTK=kq4vLaJMQtS4rf9Ag=
>>>>>> > gRbHhcBV1T2bPbp1zPZTg using mail.gmail.com>
>>>>>> > > Content-Type: text/plain; charset="utf-8"
>>>>>> > >
>>>>>> > > Dear all,
>>>>>> > >
>>>>>> > > Has any of you ever used MASEM with categorical predictors, where
>>>>>> the
>>>>>> > path
>>>>>> > > model is "X1 ~ V1 + V2 + V3", with X1 as the outcome variable and
>>>>>> V1 as a
>>>>>> > > categorical variable whilst V2 and V3 are continuous. If so,
>>>>>> could you
>>>>>> > > please point me to the paper/code? I have only found examples of
>>>>>> how to
>>>>>> > do
>>>>>> > > it with continuous predictors but never using categorical
>>>>>> variables.
>>>>>> > >
>>>>>> > > Best wishes,
>>>>>> > >
>>>>>> > > Catia
>>>>>> > >
>>>>>> > >       [[alternative HTML version deleted]]
>>>>>> > >
>>>>>> > >
>>>>>> > >
>>>>>> > >
>>>>>> > > ------------------------------
>>>>>> > >
>>>>>> > > Subject: Digest Footer
>>>>>> > >
>>>>>> > > _______________________________________________
>>>>>> > > R-sig-meta-analysis mailing list @
>>>>>> R-sig-meta-analysis using r-project.org
>>>>>> > > To manage your subscription to this mailing list, go to:
>>>>>> > > https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>>>>> > >
>>>>>> > >
>>>>>> > > ------------------------------
>>>>>> > >
>>>>>> > > End of R-sig-meta-analysis Digest, Vol 66, Issue 1
>>>>>> > > **************************************************
>>>>>> >
>>>>>> > _______________________________________________
>>>>>> > R-sig-meta-analysis mailing list @
>>>>>> R-sig-meta-analysis using r-project.org
>>>>>> > To manage your subscription to this mailing list, go to:
>>>>>> > https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>>>>> >
>>>>>>
>>>>>>         [[alternative HTML version deleted]]
>>>>>>
>>>>>> _______________________________________________
>>>>>> R-sig-meta-analysis mailing list @ R-sig-meta-analysis using r-project.org
>>>>>> To manage your subscription to this mailing list, go to:
>>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
>>>>>>
>>>>>
>>>>
>>
>>

	[[alternative HTML version deleted]]



More information about the R-sig-meta-analysis mailing list