[R-sig-eco] Good alternative to SIMPER analysis for continuous variables?

[Alessandro Manfrin]

Dear all,
   I am conducting an analysis in which I have river sites restored and
unrestored. Also, these sites are located over different altitudes,
latitudes (continuous variables) and in different fish zones (categorical
variable).

I used ADONIS (diss calculated with Bray Curtis) to assess differences in
fish trait composition between restored and unrestored sites including also
the interaction with both categorical and continuous variables:
Adonis.mod<-diss~Restoration*altitude+Restoration*latitudes+Restoration*fish
zones, strata=rivers

Now I would like to use SIMPER analysis to see which traits are responsable
for the difference observed in Restoration*fish zones (categorical
variable).

What can I use to understand which are the traits that are resonsable for
difference in Restoration*altitude as SIMPER cannot be used for continuous
variables?

Maybe stupidely I thought to do this: perform RDA (eclidean distance) on the
delta values (Rest-Unrestored) and include the significant continuous
variable from the ADONIS (i.e. altitude) in the rda analysis and then check
visually which are the "species" of traits that are responding to the vector
"altitude". Besideds the fact if it makes sense or not to do that, the other doubt
that I have with RDA is related to the "random structure=river" for which I would like
the rda to account the fact that I collected samples from different rivers in Germany (for 
which I believe I kind of considered in the Adonis with the function "strata=river"

Please tell me if I am wrong and if yes f you have some good suggestions.

Best regards

Alessandro




Dr. Alessandro Manfrin

University of Applied Sciences Trier, Umwelt Campus Birkenfeld/
University of Duisburg-Essen, Faculty of Biology

Working place:
Department of Aquatic Ecology
Room S05 T03 B02
Universitätsstr. 5
45141 Essen (DE)
Tel.: +49 (0)201/183-3113
Fax: +49 (0)201/183-2179







 

	[[alternative HTML version deleted]]