Jenny,

I suspect that many will disagree  with me about this, but here goes.

I've done several competition experiments and have never used any of 
these indices. I don't see the value in any of them. When you do your 
ANOVA, you will analyze the main effect of neighbors (competition) on 
your measure of performance. If neighbors have a significant effect, 
then you have significant interaction strength. Period. You could 
then do post-ANOVA tests and/or calculate differences between 
treatment means to get more details about what is going on. The whole 
debate several years ago about RCI vs. ACI was hogwash in my opinion. 
If the assumptions of ANOVA are met (e.g., homgeneous variance), then 
the main effect of competition (as measured by mean differences and 
associated F and p values) is a perfectly valid measure of 
competitive effect and thus competition intensity. If the assumptions 
of ANOVA are not met (e.g., heteroscedacity, which is often the case 
with non-relativized measures of performance), then you may want to 
transform the performance data.  Of course, relative growth rate is 
already a relativized measure of performance. Why does it need to be 
relativized again? If you have a multi-factor experiment, then a 
significant interaction between neighbor density and some other 
factor will tell you that interaction  strength varied with that 
factor. There is no need to relativize interaction intensity among 
treatments.

Indices are not necessary, in my opinion, unless you have 
non-relativized measures of performance, and you want to have 
standardized numbers to compare among studies. I don't see the point 
in this. I think a metaanalysis of the statistical results would be 
better. I see no need to adjust your design just to get a measure of 
statistical significance for an index of interaction intensity. I 
recommend paying more attention to the measure of performance you'll 
be using and appropriate treatment controls and seeing to it that the 
assumptions of the statistical analyses you're using are met.

That's my two cents worth.

Good luck,

Steve Brewer





At 12:39 PM -0700 7/3/07, Jenny Briggs wrote:
>Hello - a colleague and I would appreciate feedback on the question
>below. Please reply directly to:
>[EMAIL PROTECTED]
>
>Thank you very much!
>
>We are interested in using an index to measure relative interaction
>intensity.  Both Armas C.  et al. (2004) (RII) and Oksanen L. (2006)
>(CRCI=arc sin (RNE)) have proposed indices that apparently perform
>better than RCI, RNE, and lnRR.  Has anyone had experience using
>either of these or have any opinion about one over the other?  Our
>experimental design is a complete random design using an additive
>design to compare control vs. competition treatment plots.  Also,
>since it is at the community level, should we assume we should
>randomly pair up replicate control and treatment plots to calculate
>index values, so we can then run statistics?
>
>Megan Lulow
>Jenny Briggs
>
>
>--
>Jenny Briggs
>Department of Biology
>Pasadena City College
>1570 East Colorado Blvd.
>Pasadena, California 91106


-- 
Department of Biology
PO Box 1848
University of Mississippi
University, Mississippi 38677-1848

Brewer web page - http://home.olemiss.edu/~jbrewer/

FAX - 662-915-5144
Phone - 662-915-1077

Reply via email to