You are here

CI of indirect effects

5 posts / 0 new
Last post
Arin A's picture
Offline
Joined: 07/04/2016 - 05:25
CI of indirect effects
AttachmentSize
PDF icon Arin4July2016.pdf175.35 KB

Dear all,

Being very new to MetaSEM as well as R, please accept my apology for such simple question.

Mike kindly corrected my code to conduct a correlation based MetaSEM.

I just wanted to be sure that the coefficients are the standardised coefficients. Also, I was wondering whether it is possible to get the CI of the indirect effects, to know whether these effects are significant or not.
I attached the code and the results.

Thank you in advance,
Regards,
Arin

Mike Cheung's picture
Offline
Joined: 10/08/2009 - 22:37
Dear Arin, If the inputs are

Dear Arin,

If the inputs are correlation matrices, the parameter estimates are standardized.

You may obtain the 95% likelihood-based confidence intervals on the parameter estimates and the functions of the parameters, e.g., indirect effects, by adding intervals.type="LB" in tssem2(). You may check an example by calling help(Hunter83).

Best,
Mike

Arin A's picture
Offline
Joined: 07/04/2016 - 05:25
Hello Mike, Thank you very

Hello Mike,

Thank you very much for the advice.

I've included the suggestion of adding the intervals.type="LB" in tssem2() to obtain the indirect effects. However, I got NA for the SE and the Z and P values. Moreover, the OpenMx status was 6. I have attached the initial code you’ve corrected, the code where I added the intervals.type, and the respective output. I also attached the data sets from the different contexts.

I have two further questions.
1. Can I get the variance component of the random effects? (I’ve conducted a fixed effect model analysis).
2. I am now interested in conducting a Parameter-based MetaSEM on the same data sets to test the same model. I have all the necessary direct and indirect effects. I tried to follow the codes I found in the published articles (using parameter based SEM), however, I was unsuccessful in knowing how to include the direct and indirect parameters from the different studies.

I attached the two codes I tried to use: Parameter based SEM_ParameterMatrices document is a code where I tried to include the matrices of the direct effects, and Parameter based SEM_NoParameterMatrices document is a code without the matrices.

I would highly appreciate your guidance. Please do let me know if I need to upload any other information.

Thank you in advance.
Best regards,
Arin

Mike Cheung's picture
Offline
Joined: 10/08/2009 - 22:37
Hi Arin, When the LBCI

Hi Arin,

When the LBCI argument is specified, the function only returns the lower and upper LBCI. It gives NA for the SE, z and p values. It is because the SE, z and p values are based on the Wald test (SE). They may not be consistent with the LBCI. It is confusing to include both of them in the same output.

For the correlation-based analysis, the issue seems to be related to your starting values. You used 1 and 0 for the regression coefficients and error variances. It works fine when I use a different set of starting values (see the attached file).

Regarding your questions:
1. No, the heterogeneity variances are fixed at 0 in the fixed-effects model.
2. Your syntax on the parameter-based SEM is incorrect. The indirectEffect() expects a list of 3x3 correlation matrices as input. You may check the manual by calling help(“indirectEffect”). I have provided a workable model based on one of your models.

Best,
Mike

Arin A's picture
Offline
Joined: 07/04/2016 - 05:25
Dear Mike, Thank you very

Dear Mike,

Thank you very much for your prompt reply, and valuable help.

I will try to revise my code based on your suggestions.

Best regards,
Arin