R help - Jan 2012 - Null models of species co-occurrence

I am currently testing species co-occurrence patterns using null models and
the oecosimu() function within the vegan() package. My issue is that none of
the methods appear to be the ones that I want. The methods listed are r0,
r1, r2, r2dtable, swap, tswap. However, I want to know how to go about
implementing fixed row algorithms, as suggested in Gotelli 2000 in Ecology. 

Also, the null models created seem to be incredibly dependent on the 1)
burnin and 2) thin values. These are the 1) Number of null communities
discarded before proper analysis in sequential methods "swap" and
"tswap"
and 2) Number of discarded null communities between two evaluations of
nestedness statistic in sequential methods "swap" and
"tswap". What are the
significance of these values? 

--
View this message in context:
http://r.789695.n4.nabble.com/Null-models-of-species-co-occurrence-tp4325097p4325097.html
Sent from the R help mailing list archive at Nabble.com.

Dallas <tad.dallas <at> drakeresearchlab.com> writes:
> I am currently testing species co-occurrence patterns using null models and
> the oecosimu() function within the vegan() package. My issue is that none
of
> the methods appear to be the ones that I want. The methods listed are r0,
> r1, r2, r2dtable, swap, tswap. However, I want to know how to go about
> implementing fixed row algorithms, as suggested in Gotelli 2000 in Ecology.
> 
> Also, the null models created seem to be incredibly dependent on the 1)
> burnin and 2) thin values. These are the 1) Number of null communities
> discarded before proper analysis in sequential methods "swap" and
"tswap"
> and 2) Number of discarded null communities between two evaluations of
> nestedness statistic in sequential methods "swap" and
"tswap". What are the
> significance of these values? 
  You will probably have better luck with this question on the r-sig-ecology
<at> r-project.org list. 
  I haven't looked at this stuff in a little while.  It sounds from the
?oecosimu description as though there are some fixed row algorithms (?)

Methods ?r0?, ?r1? and ?r2? maintain the site (row)
     frequencies. Method ?r0? fills presences anywhere on the row with
     no respect to species (column) frequencies, ?r1? uses column
     marginal frequencies as probabilities, and ?r2? uses squared
     column sums. Methods ?r1? and ?r2? try to simulate original
     species frequencies, but they are not strictly constrained. All
     these methods are reviewed by Wright et al. (1998).

It might help in your e-mail to r-sig-ecology if you briefly
describe Gotelli 2000's algorithm, so readers don't necessarily
have to go back to the original source to know (at least approximately)
what you want.

  Ben Bolker

Dallas <tad.dallas <at> drakeresearchlab.com> writes:
> 
> I am currently testing species co-occurrence patterns using null models and
> the oecosimu() function within the vegan() package. My issue is that none
of
> the methods appear to be the ones that I want. The methods listed are r0,
> r1, r2, r2dtable, swap, tswap. However, I want to know how to go about
> implementing fixed row algorithms, as suggested in Gotelli 2000 in Ecology.
Dallas,

What do you mean with "fixed row algorithms"? If you mean the the last
row in
Table 2 of Gotelli (Ecology 81, 2606-2621; 2000) labelled "Row sums
fixed", then
these all are in vegan: SIM2 == r0, SIM4 == r1, SIM9 == quasiswap. Not all cases
of Gotelli's table are included, but could easily be added. 
> 
> Also, the null models created seem to be incredibly dependent on the 1)
> burnin and 2) thin values. These are the 1) Number of null communities
> discarded before proper analysis in sequential methods "swap" and
"tswap"
> and 2) Number of discarded null communities between two evaluations of
> nestedness statistic in sequential methods "swap" and
"tswap". What are the
> significance of these values? 
> 
The *sequential* models *are* dependent on burnin and thin: you change data very
little in one step so that matrices are dependent in the sequence. The help page
for null models contains information how to analyse this with R tools. The
easiest solution is not to use sequential models, but to use non-sequential
quasiswap. 

Like Ben Bolker wrote, you'd probably get better response in r-sig-ecology.

Cheers, Jari Oksanen