Attachment | Size |
---|---|
Final Code_July4.doc | 32.5 KB |
Figure.doc | 89.5 KB |
Output_July4.doc | 48 KB |
Dear all,
In a research project, we aim at examining a predictive model in different contexts. The model includes 9 predictors (among which 7 are mediators also), a covariate, and a dependent variable. Our model fits the four contexts very well. We would like to conduct a metaSEM which would allow us to summarise the results of our studies.
We have followed the code developed by Cheung (2016), however, our output gives us error variances which are negative, and the confidence intervals and the p-values are NA. Moreover, whenever we try to run a REM, R stops responding.
I would highly appreciate your insight as I am very new to using R. I have attached the model I am testing as well as the amended code, the output, and the data sets.
Thank you very much
Arin Ayanian
Dear Arin,
I have fixed two issues:
1) The variance of the independent variable (H) must be fixed at 1 since correlation matrices are used in the analysis;
2) When the optimization code is not either 0 or 1, it indicates that we cannot trust the results. We may try to rerun it by using the rerun() function, which uses mxTryHard() to do the hard work.
Best,
Mike
P.S. You have posted to http://openmx.psyc.virginia.edu/forums/openmx-20-discussion. It is better to post metaSEM related questions to http://openmx.psyc.virginia.edu/forums/third-party-software/metasem
Dear Mike,
Thank you very much for your prompt reply.
I highly appreciate your help.
l'll post in metaSEM forum any future inquiries I would have.
Best regards,
Arin