How to Test Mediation in AMOS (SEM)

This video explains how to test for mediation in AMOS through bootstrapping. A discussion of direct and indirect effects is presented along with step by step instructions on how to run the analysis.
If you are looking for more information about structural equation modeling, see my book titled, Applied Structural Equation Modeling using AMOS. The book covers basic topics of SEM such as confirmatory factor analysis along with path and structural models. It also covers more advanced topics such as moderation, mediation, invariance testing, and many more areas. The book can be purchased at:
Amazon:
www.amazon.com/Applied-Struct...
and
Barnes and Noble:
www.barnesandnoble.com/w/appl...

[@larissacury7714]

Amazing!! thank you!!

[@dr.kristygrayson9189]

Thanks for the great explanation. I am attempting to determine how to find in AMOS or calculate the t-value you have in noted parentheses on your template table. Would you have a quick explanation? Thanks!

[@joelcollier9387]

In the output in the estimates tab, you will see a column called "C.R". That stands for Critical ratio which is also the T-value for that specific relationship you are trying to test. Hope that helps.

[@muti7880]

thanks for the video Dr. I wonder do you have a tutorial video on what data is inserted on the model? Like tutorial on how you compute your data in spss.

[@joelcollier9387]

I don't have a video on it yet but that is a good idea. In my book Applied Structural Equation Modeling using AMOS, I give alot of space to cleaning data and running the analysis in SPSS before moving to AMOS. That might be an option in the meantime.

[@talzabidi1569]

Thanks Dr 1-what is the different between mediation and indirect effect ? 2-C.R is express t-value or z value ? because when I click on it they said Z value? 3- why we only look to unstandardized estimate?

[@joelcollier9387]

mediation is the process of examining the indirect effect from the IV to the ultimate DV. C.R. stands for critical ratio and it is the t-value. We only look at unstandardized because of the bootstrap. Standardizing a bootstrap can produce estimates that are slightly off because of the standardization.

[@talzabidi1569]

But C.R in AMOS output when I click in the value it shows Z-value Not t-value, could you please explain it?

[@DK-em6oz]

Hi Collier, thanks for this very informative videos. I am facing an issue in my model: my Amos does not bring out the Estimate/Bootstrap results window for me to check the two tail significance (BC). Is there anything I am doing wrong? My measurement model is a second-order model. This happens when I add control variables to the model

[@joelcollier9387]

Amos is a little quirky. Once you run the analysis and go into the results, you need to click on the "matrices" link and then click on "indirect effects". Once you do that, then you should see the links at the bottom titled "Bootstrap Confidence" become active. Clicking in that link with show you the confidence intervals. You have to click on the link of indirect effects on matrices before the confidence interval info will be active. Hope this helps.

[@DK-em6oz]

@@joelcollier9387 Thanks Collier, Some news about the progress: strangely the bootstrap results window disappears when I add controls to the model with arrows from the controls to both the mediator and the DV. When the arrows move only go to the DV suddenly bootstrap results boom! show up. I also face this new problem: when I bootstrap 2000 samples in the mediation analysis I receive an error msg that "During the analysis of a bootstrap sample, an attempt was made to compute a standardized regression weight between two variables, one of whose estimate failed to be positive. The attempt was made because ' Standardized Estimates' in the 'analysis properties' window was checked or because the Standardized method was used." when I bootstrap 30 samples the analysis runs just fine. Have you encountered this before? How may I overcome this if I want over 500 bootstrap samples?

[@joelcollier9387]

@@DK-em6oz Ok. That makes more sense now. First, you are only concerned with unstandardized results with a bootstrap. If you have constructs that are strongly correlated with one another (maybe your control variables) selecting the option of presenting the standardized regression weights will bomb out the program because standardizing the data produces values over 1 (its a multicollinearity issue). Go in the output and uncheck provide standardized estimates. Now run the analysis with the control variables and it should present all the results. Again, with a bootstrap you only want the unstandardized results anyway.

[@DK-em6oz]

@@joelcollier9387 And just with a click the model worked perfectly!!! Thank you so much, I would NEVER have figured out this by myself.

[@muhammadaamirkhan8182]

If the IV to Mediator path is significant and; Mediator to DV path is significant; then, is it necessary that the indirect effect of IV to DV via Mediator MUST be significant?