[R-sig-eco] Multivariate ANOVA based on dissimilarities

[Jari Oksanen]

Mauricio,

Nothing reproducible here (we don't have your data sets), not even the output.

In your email you only write about adonis() and the p-values you got from adonis. On the other hand, your TukeyHSD() is based on betadisper(), and you never seem to check the betadisper() p-values. These two methods are radically different: adonis() is about differences between groups, and betadisper() is about dispersions within groups. We expect some kind of consistency between betadisper() p-values and TukeyHSD() of betadisper, but adonis() is quite a different beast which has little to do with betadisper().

If you take an analogy to standard statistics: adonis() is like (M)ANOVA, betadisper() is like Levene's test of "homogeneity" (what ever that means) of variances, and TukeyHSD() is its sequel about detecting cases with different variances.

I suggest you rather use 999 permutations than 1000: you can get nicer round figures. Once we were more patronizing and dropped one permutation if the user asked for even hundred or thousand, but then we thought that this is up to user to decide and none of our business.

Cheers, Jari Oksanen

On 27/08/2012, at 23:23 PM, Maurício Moraes Zenker wrote:

> Dear all,
> 
> 
> I‚m using the function „adonis‰ (multivariate ANOVA based on
> dissimilarities) to study beta diversity of moths between altitudinal
> levels and seasons. My analysis is based on the example included in the
> vegan tutorial „Multivariate Analysis of Ecological Communities in R: vegan
> tutorial‰, written by Jari Oksanen. As in the example, the analysis of my
> data indicated that there are significant differences. However, the Tukey‚s
> test indicated that there are no differences between groups? Why does it
> happen? Shouldn‚t the Tukey‚s test show that one group is different from
> another?
> 
> The script is almost the same of the tutorial but with more permutations.
> 
> 
> 
> Thank you.
> 
> 
> 
> Mauricio
> 
> 
> 
> Permanova<-read.table("Permanova.txt",header=T)
> 
> PermanovaII<-read.table("PermanovaII.txt",header=T)
> 
> betad <- betadiver(Permanova, "z")
> 
> adonis(formula = betad ~ Altitude*Season,data = PermanovaII,
> 
> permutations = 1000)
> 
> hov <- with(PermanovaII, betadisper(betad, Altitude))
> 
> TukeyHSD(hov)
> 
> hovII <- with(PermanovaII, betadisper(betad, Season))
> 
> TukeyHSD(hovII)
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

-- 
Jari Oksanen, Dept Biology, Univ Oulu, 90014 Finland
jari.oksanen at oulu.fi, Ph. +358 400 408593, http://cc.oulu.fi/~jarioksa