SCM

Forum: help

Monitor Forum | Start New Thread Start New Thread
Monte Carlo tests for subsequent axis of PRC [ reply ]
By: Iker Dobarro on 2012-11-22 12:04
[forum:34930]
Hello to all. I'll try to be concise. I've been exploring funtion prc and I've been trying to test axis significance after the first one. I've data analysed with CANOCO v4.53 and I've been comparing with it the results from vegan. Using the parameter by="axis" in anova.cca, I obtained very different results with CANOCO ones -up to 5 significative axis, when CANOCO gives me only the first one-.

It seems that parameter by="axis" doesn't take the correct scores from the prc analysis as covariables for the Monte Carlo test. I think this because I searched for the so-called "SamE" CANOCO's scores in prc results, which seem to be cca.object$CCA$u. I extracted these scores to a new vector and I run a rda with this vector as Condition, what, I think, is the method suggested by Leps and Smilauer:

secondaxis <- rda (species ~ Treatment * Time + Condition (Time) + Condition (SamE)

After that, I analysed for the first axis of this last rda with anova.cca, which is supposed to be de 2nd axis of the former prc analysis. Then, I obtained a result similar to CANOCO's. Although, in some ocassions, with other data, eigenvalues given by prc and later rda varies a little bit.

I don't know which causes these differences, if I'm doing well with extracting sample scores which are linear combinations of environmental variables (SamE) for doing the analysis or if it's simply a bug in by="axis" parameter.

Thanks for your attentions, and specially to Dr. Oksanen for his effort.

Thanks to:
Vienna University of Economics and Business University of Wisconsin - Madison Powered By FusionForge