How to test for linear effects?

[d***@gmail.com]

Hi everyone,

I performed a mixed ANOVA where I had 7 intensity levels as within-factor and age (old-young) as between factor. We then used multiple comparison (corrected with Bonferroni) to test the different intensity levels (when significant).

Now one of our reviewers is asking to test for significant linear effects and potential interaction of the linear effect with age to gain statistical power and to avoid making so many comparisons.

I'm really lost here, any ideas on how to do this?

Thanks!
Diana

[Bruce Weaver]

Post by d***@gmail.com
Hi everyone,
I performed a mixed ANOVA where I had 7 intensity levels as within-factor and age (old-young) as between factor. We then used multiple comparison (corrected with Bonferroni) to test the different intensity levels (when significant).
Now one of our reviewers is asking to test for significant linear effects and potential interaction of the linear effect with age to gain statistical power and to avoid making so many comparisons.
I'm really lost here, any ideas on how to do this?
Thanks!
Diana

Did you use GLM REPEATED MEASURES (Analyze > General linear model > Repeated measures in the GUI)? If so, your command syntax will look something like this example from the second link below (but with 7 levels of the within-Ss factor):

GLM Y1 TO Y4 BY GROUP
/WSFACTOR=YEAR 4 POLYNOMIAL
/WSDESIGN=YEAR
/PRINT=PARAMETER
/DESIGN=GROUP.

To get polynomial contrasts for GROUP (the bewteen-Ss variable), add a CONTRAST sub-command, like this:

GLM Y1 TO Y4 BY GROUP
/WSFACTOR=YEAR 4 POLYNOMIAL
/CONTRAST(race)=Polynomial /* <<<< ADD THIS */
/WSDESIGN=YEAR
/PRINT=PARAMETER
/DESIGN=GROUP.

This can also be done via the GUI by clicking on the Contrasts button.

[d***@gmail.com]

Thanks Bruce, I got the contrasts! I thought the reviewer wanted something else, it's so weird that for once they want something simple!

Diana

Post by Bruce Weaver

Post by d***@gmail.com
Hi everyone,
I performed a mixed ANOVA where I had 7 intensity levels as within-factor and age (old-young) as between factor. We then used multiple comparison (corrected with Bonferroni) to test the different intensity levels (when significant).
Now one of our reviewers is asking to test for significant linear effects and potential interaction of the linear effect with age to gain statistical power and to avoid making so many comparisons.
I'm really lost here, any ideas on how to do this?
Thanks!
Diana

GLM Y1 TO Y4 BY GROUP
/WSFACTOR=YEAR 4 POLYNOMIAL
/WSDESIGN=YEAR
/PRINT=PARAMETER
/DESIGN=GROUP.
GLM Y1 TO Y4 BY GROUP
/WSFACTOR=YEAR 4 POLYNOMIAL
/CONTRAST(race)=Polynomial /* <<<< ADD THIS */
/WSDESIGN=YEAR
/PRINT=PARAMETER
/DESIGN=GROUP.
This can also be done via the GUI by clicking on the Contrasts button.

[Bruce Weaver]

Post by d***@gmail.com
Thanks Bruce, I got the contrasts! I thought the reviewer wanted something else, it's so weird that for once they want something simple!
Diana

No worries. And for the benefit of those who read the archives later, the example I gave should have had GROUP (not RACE) as the variable in the CONTRAST subcommand.

GLM Y1 TO Y4 BY GROUP
/WSFACTOR=YEAR 4 POLYNOMIAL
/CONTRAST(GROUP)=Polynomial /* <<<< ADD THIS */
/WSDESIGN=YEAR
/PRINT=PARAMETER
/DESIGN=GROUP.