Different results when using Mediation and SEM Module in JASP

Miriam__

Hi JASP-ers,

I am currently conducting some mediation analyses for my master thesis. Unfortunately, I do not get the same results when using the Structural Equation Modeling and the Mediation Analysis under the tab SEM in JASP. I copied the lavaan syntax from the mediation calculation and entered it into my SEM model. With my SEM, I don't get a significant indirect effect, however, when using mediation analysis, that effect is highly significant. Could it be that because I am using variables with measurement error included in mediation, mediation gets me different results there? I did not see an option in JASP that lets me calculate my factor scores from my latent variables calculated with SEM :(

I've attached a pdf with my results that may make it a bit clearer :)

Thankful for every help and answer!! :)

SEM.pdf

Best

Miriam

evankesteren

Hi Miriam,

that's a very interesting difference, because not only the standard errors are different (which could potentially be due to choosing a different standard error method, bootstrap vs. normal), but also the parameter estimates (for example the total effect). If you are using different variables for the two, you will indeed get different results and the difference could be from this.

Unfortunately, without knowing more about the specific model and settings, I cannot say anything about where this difference comes from. If I copy the lavaan syntax from mediation to the SEM module and use the same dataset, for me it shows the same results.

Which version of JASP are you using? Are you able to share a JASP file? If not publicly, you may send it to ej.vankesteren@jasp-stats.org

Erik-Jan

meredhian

This also happened to me, can you explain whyy? :(

EJ

I think the same advice holds here as well: you could send Erik-Jan an email. Of course it would be great if he could share the "solution" on this forum to avoid getting these emails on a regular basis :-)