Hi there!
The figure shows a parameter I am estimating for each species present in a site. Blue dots are replicates obtained by bootstrapping the site by species matrix. Red squares are median values and red lines indicate the empirical 95% CI intervals. Each plot (A-C) represent a replicate from a given locality. I have in all 5 localities to compare. I would like to find out the amount of variation among localities and among replicates within localities. In particular, is the variation among replicates comparable to variation among localities?
Note that not all species are present in each plot, but their values can be considered to be zero.
Any idea will be most welcome.
Thank you very much for your time.
Gracias Jorge. I will have a look at it. At first sight it looks more like testing effects, whereas I am more interested in partitioning of variation. Anyway, it will serve as starting point. Thanks
Looking at the bootstrapped data it appears there are multiple distributions, which is why, I assume you used median and empirical 95% CI. There are also big differences among species in terms of occurrence. Maybe by dual scaling your 'global' matrix (i.e., matrix of all data sums to 1) you can look at the overall and then from there partition the variance into site, species, and location.
For some reason the attached paper comes to mind, in terms of looking at underlying hypotheses for community structure. I used something similar for diet overlap among morphologically similar and related fish species.
Thank you very much Patrick. I will read Lawlor (1980) with great attention.
Is there a specific reason for dual scaling the global matrix? My data ranges from 0-1. So everything is measured on the same scale. Just wondering.
I was thinking you could apply the basic idea of a correspondence analysis, where each 'presence' in the global matrix would be equivalent in terms of its contribution (i.e., F.. = 1) and linearly additive... just a thought.
I think I may not need the bootstrapped data. I will probably run varpart (vegan in R) to assess the variation accounted by site and time. The bootstrapped data might still be used to compute CIs for the variation.
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix