Re: How to nullified the correlation

this is also called a repeated measures.
It is also called a 2 by 2R design, or a 2 by 2 design with one between groups factor and one within group (repeated measure).

You don't want to nullify the inclusion correlation of the measures. You want to take advantage of it.

What are the details on how your DV is measured? Are the intervals "equal appearing"? Are they very discrepant from interval?

One common approach would be to do the ANOVA on cataract level as is and then rank all of the measures of level and run the ANOVA on ranks. See if your substantive conclusions differ. If the conclusions are similar, breathe a sigh of relief.

Since the levels for the two eyes are positively correlated, you could throw away power and do the anova on the average level.

Another more complex approach would be to see if a categorical regression (CATREG) would run using a set of dummy variables (N-1) to control for subject. (not something I have done, but this would be the way to do a 2 *2R anova via a regression procedure.)

Art
Art@xxxxxxxxxxxxx
Social Research Consultants


RAMS wrote: 
I have an ordinal value for the cataract level.  So i cant perform
ANOVA.  What is the alternative in a Non-Parametric technique?

.

Relevant Pages

  • Re: How to nullified the correlation
    ... You don't want to nullify the inclusion correlation of the measures. ... then rank all of the measures of level and run the ANOVA on ranks. ... If you are going to post to multiple newsgroups or discussions lists, please use multiple addresses on the same post. ...
    (sci.stat.math)
  • Re: Type of ANOVA model
    ... How would one perform rank based version of a repeated measures ANOVA ... in SAS? ... Bruce Weaver ...
    (sci.stat.consult)