Non-parametric Multivariate Analysis

#1
Hi everybody,

Is there anything like a non-parametric multivariate test?
I would like to analyse whether the proportion of visited flowers per plant is affected by the presence of another plant and the type (e.g. species) of insect visitor.

With a Kruskal-Wallis test I can either test the one or the other. I am also interested in the interaction between the factors.
Can anyone help?

Many thanks,


Thomas
 
#2
The question is why do you need a non-parametric test? Have you already tried a "standard" multivariate model (anova or regression) and found that it violates some assumptions?

You will likely need a bootstrap or some other simulation method. It sounds like it could be complicated, too, but not impossible.

Dave
 
#3
Hi everybody,

Is there anything like a non-parametric multivariate test?
I would like to analyse whether the proportion of visited flowers per plant is affected by the presence of another plant and the type (e.g. species) of insect visitor.

With a Kruskal-Wallis test I can either test the one or the other. I am also interested in the interaction between the factors.
Can anyone help?

Many thanks,


Thomas
Sure there is, take a look to PERMANOVA software provided for Marti Anderson. It is similar to ANOSIM analysis that you can perform with a software called PRIMER. I attached both, software and user notes. First read carefully the user notes and let me know if is useful for you.

p.s.: actually this is a semi-parametric analysis, but is really robust again statistical errors. Of course if you have any question more, let me know

Salu2