Hi All,

Thanks in advance for your help and time! It is much appreciated.

I have 11 datasets (15 variables: 3 latent/3 indicators each, 2 observed DVs, 4 control variables) and am interested in using TSSEM. However, I'm running into errors and am getting stuck at stage 1. I have uploaded my data, syntax, and output.

For stage 1, I first started off with the fixed effects model and decided that, based on the output, an assumption of homogeneity was not reasonable given the results. I then tried running a random-effects model but received error code 6. I tried following the suggestion given on the error page but got an error message stating that "mxRun can only act on MxModel objects." I'm not quite sure how to proceed from here and am uncertain if it's an issue with my syntax, data, or model.

Additionally, I was had some questions about the second stage. Most of the examples I found illustrated how to set up the matrices and do the analyses for models that looked at the correlations between latent variables. However, for my stage two model, I would like to test a model that includes indirect effects between some of the latent variables, regressions from the latent variables to observed DVs, and control variables. Are there any example syntax available or resources that can help guide me through how to do this?

Finally, is there any way to test for the influence of dataset level variables (e.g., publication year, type of measurement used) outside of using subgroup analysis?

Sincerely,

Anne

Hi Anne,

For some reason, there are strange symbols in your Syntax.txt.

The analysis is very challenging because you are estimating the random effects with only 10 studies. In some correlation coefficients, there are only 6 studies.

There was a syntax error in

It should be

Regarding the Stage 2 analysis, you may use the lavaan model specification and convert it to RAM specification. See examples in https://cran.r-project.org/web/packages/metaSEM/vignettes/Examples.html#two-stage-structural-equation-modeling-tssem

Continous moderators may now be used to moderate the parameters in MASEM. See https://psyarxiv.com/ce85j/ and https://github.com/mikewlcheung/code-in-articles/tree/master/Jak%20and%20Cheung%202019

I hope it helps.

Mike

Dear Dr. Cheung,

Thank you for your help and the resources! I have fixed the syntax error and did both the firstrun and secondrun. However, the error code was not resolved; I have attached the output and the fixed syntax below. Are there any alternative solutions that I can try? Thank you!

Hi Anne,

The problem is that there are 210 parameters (105 average correlations and 105 variances). However, there are only 10 or fewer rows of data. It is difficult to justify a simple meta-analysis with 10 studies. You will need much more data to conduct a MASEM.

Mike

Thank you for your help!