Surrogacy1.1 applied on a Coral Reef Island

[Mayeul Dalleau]

Dear community,

I previously tried to contact Dr Garson and Dr Sarkar using
con...@uts.cc.utexas.edu and I've been redirected here. So I'm
taking
this opportunity to contact you about my issues.

I'm currently testing the surrogacy of benthic habitats against
species
richness on the coral reef island of Wallis (South-Pacific). I'm
using
the Surrogacy 1.1 software that I downloaded from
http://uts.cc.utexas.edu/~consbio/Cons/ResNet.html . By now,
I obtained some very interesting results but one came out which I can
not really explain.

Let's say I got 42 sites, each of which is associated (1) with a
'presence/absence' list of benthic habitats (49 in total) present in a
given environment and (2) with a 'presence/absence' list of species
(151
in total) observed on the field. When I choose a target of 100% for
the
estimator surrogate (namely the benthic habitats), the algorithm
should
not stop till all sites are selected (in this case 42 sites). The
proportion of true surrogate (namely the species) that met their
target
should then be of 100% too (I chose a target of 1 for the true
surrogate, but the result I'm talking about should not depend of this
target). This is actually what I get using Surrogacy 1.1. Now, with
the
same target for the estimator surrogate, the random simulation (1000
repeats) should give as well a mean of 100% species that have met
their
target since all 42 sites are selected (standard error should be equal
to 0). Surprisingly, I do get a mean of 88.17% with a really small (no
to say null) standard error. How is this possible? I checked my data
several times and tried some other data but still I do get this kind
of
result. One justification, according to me, would be that the random
method samples the sites with replacement. That would explain such a
result. Is it really the case ? If yes, why is it so ?

I uploaded to this group the settings, the input files and screenshots
of the graphs to illustrate my words.

I would really appreciate your comments on that subject and I'm open
for
a possible future collaboration.

With all my regards,

Mr Mayeul Dalleau
Institut de Recherche et de Développement
Noumea, New Caledonia

[Sahotra Sarkar]

In the first part of your analysis the program did exactly what it was
supposed to (and what you expected) but with one caveat: what does it
mean that you used true surrogates for this run? In the second part,
it is not clear to me what you did (though it is true that the random
sampling does use sampling with replacement). If you set the 100 %
target, that would be the target for estimator surrogates. Then, what
would be checked is the extent to which the true surrogates met their
targets. This can be any number less than 100 %. So, I am not sure
what else you should be expecting.

On Mar 26, 5:05 pm, Mayeul Dalleau <mayeul.dall...@gmail.com> wrote:
> Dear community,
>
> I previously tried to contact Dr Garson and Dr Sarkar using

> cons...@uts.cc.utexas.edu and I've been redirected here. So I'm

> taking
> this opportunity to contact you about my issues.
>
> I'm currently testing the surrogacy of  benthic habitats against
> species
> richness on the coral reef island of Wallis (South-Pacific). I'm
> using

> the Surrogacy 1.1 software that I downloaded fromhttp://uts.cc.utexas.edu/~consbio/Cons/ResNet.html. By now,