Mauchly's sphericity test

[Michelle Lim]

Hi,
I am running analysis to find out between and within group (pre-post) differences for an intervention study. It is based on 2x measurement pre-post with multiple outcome measures. I used repeated measure on the spss but I am not sure why does the analysis reflect empty under the sig column for the mauchly's sphericity test. Since it is empty, is it still possible to use the results reflected in the multivariate table?

Thank you!

With regards,
Michelle

[SR Millis-3]

Michelle,

 

You have only 2 time periods and, thus, a single variance.  As you know, sphericity relates to the equality of the variance of the differences between levels of the repeated measures factor.

 

With a pre-test, post-test design, you are better off using an analysis of covariance with the pre-test score as the covariate.

 

~~~~~~~~~~~
Scott R Millis, PhD, ABPP, CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [hidden email]
Email: [hidden email]
Tel: [phone]
Fax: [phone]

--- On Tue, 6/22/10, Michelle Lim <[hidden email]> wrote:

From: Michelle Lim <[hidden email]>
Subject: Mauchly's sphericity test
To: [hidden email]
Date: Tuesday, June 22, 2010, 11:24 AM

Hi,
I am running analysis to find out between and within group (pre-post) differences for an intervention study. It is based on 2x measurement pre-post with multiple outcome measures. I used repeated measure on the spss but I am not sure why does the analysis reflect empty under the sig column for the mauchly's sphericity test. Since it is empty, is it still possible to use the results reflected in the multivariate table?

Thank you!

With regards,
Michelle

[Bruce Weaver]

To complete the thought in Scott's first paragraph below, if there are only two levels, there is only one difference, so there cannot be heterogeneity of variance of the paired differences.  THAT is why you are not getting anything for Mauchly's test of sphericity.  You should also see, in this case, that epsilon = 1.

FWIW, I agree with Scott's suggestion that ANCOVA with pre-test score as the covariate is preferred for this design.

You also mentioned that there are multiple outcome measures.  How many, exactly?  Can one or two of them be classified as the primary outcomes, with the rest considered secondary?   What are the sample sizes for your two groups?  What I am getting at is that if you carry out enough tests, some will be statistically significant by chance.  When you have multiple outcomes, you must take some steps to control for this.

Bruce

SR Millis wrote

Michelle,
 
You have only 2 time periods and, thus, a single variance.  As you know, sphericity relates to the equality of the variance of the differences between levels of the repeated measures factor.
 
With a pre-test, post-test design, you are better off using an analysis of covariance with the pre-test score as the covariate.
 
~~~~~~~~~~~
Scott R Millis, PhD, ABPP, CStat, CSci
Professor & Director of Research
Dept of Physical Medicine & Rehabilitation
Dept of Emergency Medicine
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: [e-mail]
Email: [e-mail]
Tel: [phone]
Fax: [phone]

--- On Tue, 6/22/10, Michelle Lim <[e-mail]> wrote:


From: Michelle Lim <[e-mail]>
Subject: Mauchly's sphericity test
To: [e-mail]
Date: Tuesday, June 22, 2010, 11:24 AM


Hi,
I am running analysis to find out between and within group (pre-post) differences for an intervention study. It is based on 2x measurement pre-post with multiple outcome measures. I used repeated measure on the spss but I am not sure why does the analysis reflect empty under the sig column for the mauchly's sphericity test. Since it is empty, is it still possible to use the results reflected in the multivariate table?

Thank you!

With regards,
Michelle