Scribd
Upload
190 views15 pages

Bootstrapping in Amos: Changya Hu, Ph.D. NCCU

Bootstrapping is a resampling method. Creating an sampling distribution to estimate standard errors, and create the confidence intervals. It's important for mediation analysis Because of its accuracy for computing confidence intervals for mediation effect when the mediation effect is nonzero.

Uploaded by

Chamid Mahfudz
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
190 views15 pages

Bootstrapping in Amos: Changya Hu, Ph.D. NCCU

Bootstrapping is a resampling method. Creating an sampling distribution to estimate standard errors, and create the confidence intervals. It's important for mediation analysis Because of its accuracy for computing confidence intervals for mediation effect when the mediation effect is nonzero.

Uploaded by

Chamid Mahfudz
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Bootstrapping in Amos

Changya Hu, Ph.D. NCCU

Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

Whats bootstrapping and why we need it?


Its a resampling method
Creating an sampling distribution to estimate standard errors, and create the confidence intervals.

Its important for mediation analysis


To confirm the mediation effect
Because of its accuracy for computing confidence intervals for mediation effect when the mediation effect is nonzero.

As an aid to nonnormal data


The assumption of SEM is the data has a multivariate normal distribution, but many empirical studies failed. The resampling method has more accurate Type I error rates and power than single sample method that assumes a normal distribution.
Page 2 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

Benefits of bootstrap procedure


SEM approach
If the variables have measurement errors, the significance of the mediation effect is likely to be underestimated.
Using SEM can deal with the measurement error problem.

Benefits
It allows researchers to assess the stability of parameter estimates It can be applied when the assumptions of large sample size and multivariate normality may not hold.
Needs at least moderate sample sizes

Page 3 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The procedure of bootstrapping in Amos

Click on the Analysis Properties button

And click on the Output and Bootstrapping tab

Page 4 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The procedure of bootstrapping in Amos

To test the mediation effect

Page 5 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The procedure of bootstrapping in Amos

Usually 500 or 1,000 (Cheung & Lau, 2008)

Determine the Type I error rate

Page 6 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos

Page 7 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos

The original ML estimates No mediation effect


Page 8 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos

The Bootstrap estimates Showing the S.E.


Page 9 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos

The CIs We can find all the bootstrapped results of the estimates from this field
Page 10 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos


How to check the mediation effect?

Page 11 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos

Indirect effect of depress on selfest is: .011 (95% CI: -.005 ~ .106)
Page 12 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

The result of bootstrapping in Amos


Indirect effect of depress on selfest is: .011 (95% CI: -.005 ~ .106)

X1

21
X2

Y1.2
Y

Y2.1

We cannot reject the null hypothesis, H0: 21 * Y2.1 = 0 There is no mediation effect! (Cheung & Lau, 2008)
Page 13 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

References
Cheung, G. W., & Lau, R. S. (2008). Testing mediation and suppression effects of latent variables: Bootstrapping with structural equation models. Organizational Research Methods, 11(2), 296-325. MacKinnon, D. P. (2008). Introduction to Statistical Mediation Analysis. New York, NY: Lawrence Erlbaum Associates. Byrne, B. M. (2001). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Mahwah, NJ: Lawrence Erlbaum Associates.
Page 14 Copyright 2010. Changya Hu & Yu-Hsuan Wang. All rights reserved

You might also like