[R-sig-eco] mvabund: Extracting a subset of comparisons between interactions

[Dixon, Philip M [STAT]]

Diogo,

In similar situations, I have done the following:
Estimate all pairwise differences between site:season combinations.  Use paste() or something equivalent so that the group means encode both site and season.  No p-value adjustment at this step.
Post-process that table to extract only the comparisons of interest, e.g. comparisons of season means within site or comparison of site means with season.
Determine the number of comparisons you are interested in (likely much, much smaller than 498).  Apply a Bonferroni adjustment to your subset of comparisons.

I don't have R code to do this with anova.manyglm(  pairwise.com= ) output.  Sorry.  Here's an outline of what I would do:
If you save the results from anova.manyglm(), the information you need to extract comparisons of interest will be in the row dimnames of $pairwise.comp.table.
E.g.:

all.pairs <- anova.manyglm(full_mod, p.uni="adjusted , pairwise.comp = Catch$SiteSeason)
pair.names <- dimnames( all.pairs$pairwise.comp.table)[[1]]

Then it's just a matter of using strsplit() three times.  Once to separate the two group names (using ' vs ').  Then to each group name to separate the site from the season (using '_' or whatever you used as the separator character in your paste).
Best,
Philip Dixon


	[[alternative HTML version deleted]]