Hi, thanks for your interest in CEM!
We are transitioning the CEM email list to GitHub.
If you have a bug or feature request, please click
https://github.com/IQSS/cem/issues.
If you have a general question, comment, or suggestion, click
https://github.com/IQSS/cem/discussions
Thanks again,
Stefano, Gary, and Beppe
Hello!
I had some questions in trying to apply CEM to mediation analysis.
SEM methods work fine with the CEM weights so that on the surface, mediation seems easy. However, I am not sure that the standard errors are correct. In mediation analyses, it is recommended to do bootstrapping instead of using the output z-values and corresponding p-values as mediated variables often do not follow normal distributions to be highly conservative.
I have thought of doing bootstrapping with the weights just on the matched sample but am not if this is the correct approach. For example,
https://www.stata.com/statalist/archive/2012-02/msg00822.html suggests that bootstrapping needs to be done more complicatedly in order to correctly incorporate variability.
In sum, I would like to get some feedback on how to correctly do mediation (and mediated moderation) analysis using CEM. Even if complete answers are not possible, I would still truly appreciate any kind of related information.
Thank you in advance.
Best,
Yon Soo Suh
CEM is not designed primarily for mediation analysis, but you can do CEM to
reduce model dependence at the first stage and then do whatever type of
mediation or other analysis you might do afterwards.
best of luck with your research,
Gary
--
*Gary King* - Albert J. Weatherhead III University Professor - Director,
IQSS <http://iq.harvard.edu/> - Harvard University
GaryKing.org <http://garyking.org/> - King(a)Harvard.edu - @KingGary
<https://twitter.com/kinggary> - 617-500-7570 - Assistant
<king-assist(a)iq.harvard.edu>: 617-495-9271
On Thu, Dec 3, 2020 at 3:48 AM 서연수 <yssshinhwa(a)naver.com> wrote:
> Hello!
>
> I had some questions in trying to apply CEM to mediation analysis.
>
> SEM methods work fine with the CEM weights so that on the surface,
> mediation seems easy. However, I am not sure that the standard errors are
> correct. In mediation analyses, it is recommended to do bootstrapping
> instead of using the output z-values and corresponding p-values as mediated
> variables often do not follow normal distributions to be highly
> conservative.
>
> I have thought of doing bootstrapping with the weights just on the matched
> sample but am not if this is the correct approach. For example,
>
> https://www.stata.com/statalist/archive/2012-02/msg00822.html suggests
> that bootstrapping needs to be done more complicatedly in order to
> correctly incorporate variability.
>
> In sum, I would like to get some feedback on how to correctly do mediation
> (and mediated moderation) analysis using CEM. Even if complete answers are
> not possible, I would still truly appreciate any kind of related
> information.
>
> Thank you in advance.
>
>
>
> Best,
>
>
>
> Yon Soo Suh
>
>
>
