Different results in unmarked and PRESENCE for Royle model

134 views
Skip to first unread message

Manuel Spínola

unread,
Jul 10, 2011, 8:44:56 AM7/10/11
to unma...@googlegroups.com
Dear list members,

I got different results using unmarked and PRESENCE, and I would like to know why.

Output from PRESENCE

 ********* Input Data summary *******
Number of sites                = 60
Number of sampling occasions   = 5
Number of missing observations = 0
Data checksum = 46204

Naive occupancy estimate       = 0.8833

NSiteCovs-->3
site_covname[0]=Río1
site_covname[1]=Río2
site_covname[2]=Río3
NSampCovs-->0

-->5-1
Matrix 1: rows=3, cols=5
-,a1,a2,a3,a4,
p  1 0 0 0
lambda  0 1 Río2 Río3

Royle Model w/ species counts (k=200)
Number of groups               = 1
Number of sites                = 60
Number of sampling occasions   = 5
Number of missing observations = 0


Number of parameters           = 4
Number of function calls       = 121
Final function value           = 539.181500
-2log(likelihood)              = 1078.363000
AIC                            = 1086.363000
Naive occupancy estimate       = -1.#IND00

Untransformed (beta) parameters:
Estimated parameter                  estimate  std.err
--------------------------           --------  -------
p                                    = -3.4206   0.0688
lambda                               = 3.4615   0.1248
lambdaRío2                           = 4.3267   31622776601.6838
lambdaRío3                           = -0.5095   0.1724

beta var-cov matrix:
    0.0047   -0.0046    0.0000    0.0000
   -0.0046    0.0156    0.0000   -0.0112
    0.0000    0.0000 999999999999999870000.0000    0.0000
    0.0000   -0.0112    0.0000    0.0297

Output from unmarked

> summary(fm.ranas)

Call:
pcount(formula = ~1 ~ x2, data = r, K = 30, mixture = "P")

Abundance (log-scale):
            Estimate    SE     z  P(>|z|)
(Intercept)    2.219 0.143 15.47 5.16e-54
x2b            0.882 0.138  6.39 1.64e-10
x2c           -0.491 0.188 -2.61 9.00e-03

Detection (logit-scale):
 Estimate     SE     z  P(>|z|)
    -2.07 0.0983 -21.1 1.95e-98

AIC: 1106.519 
Number of sites: 60
optim convergence code: 0
optim iterations: 52 
Bootstrap iterations: 0 


Thank you very much in advance.

Best,

Manuel

--
Manuel Spínola, Ph.D.
Instituto Internacional en Conservación y Manejo de Vida Silvestre
Universidad Nacional
Apartado 1350-3000
Heredia
COSTA RICA
mspi...@una.ac.cr
mspin...@gmail.com
Teléfono: (506) 2277-3598
Fax: (506) 2237-7036
Personal website: Lobito de río
Institutional website: ICOMVIS

Andy Royle

unread,
Jul 10, 2011, 10:24:54 AM7/10/11
to unma...@googlegroups.com, unma...@googlegroups.com
hi Manuel,
 there's no telling what the problem is here.
 For one thing you could check whether "K" has an effect on things by increasing its value.
 
 Secondly, I suggest simulating a data set exactly like yours (using the parameter values estimated from unmarked) and then fit that data set in unmarked and PRESENCE. That should diagnose if one or the other is in error.
 
 Of course we would all be interested to see the result!
andy

 
J. Andy Royle
Research Statistician
USGS Patuxent Wildlife Research Center
12100 Beech Forest Rd.
Laurel, MD 20708
http://profile.usgs.gov/professional/mypage.php?name=aroyle
andy_...@usgs.gov
phone: 301-497-5846
fax: 301-497-5545

Book: "Hierarchical Modeling and Inference in Ecology: The Analysis of Data from Populations, Metapopulations and Communities" by J. A. Royle and R.M. Dorazio.  

unmarked: A very useful R package for fitting certain hierarchical models using likelihood methods. Available from: http://cran.case.edu/web/packages/unmarked/index.html

-----unma...@googlegroups.com wrote: -----

To: unma...@googlegroups.com
From: Manuel Spínola <mspin...@gmail.com>
Sent by: unma...@googlegroups.com
Date: 07/10/2011 08:44AM
Subject: [unmarked] Different results in unmarked and PRESENCE for Royle model

Richard Chandler

unread,
Jul 12, 2011, 7:50:01 AM7/12/11
to unma...@googlegroups.com
Hi Manuel, 

I just got the same results when fitting the model to the data used in the "Class6.R" script found here: https://sites.google.com/site/hierarchicalmodelingcourse/home/r-scripts. 


PRESENCE:
                                            estimate   std.error
A1     : p                                  0.157462 (0.085572)
A2     : lambda                            -0.172146 (0.080491)


unmarked
                Est         SE
p(Int)    0.1574618 0.08557164
lam(Int) -0.1721461 0.08049056


So, I can't replicate the problem. Did you by chance standardize covariates in unmarked, but not in PRESENCE? You should also use the same K as Andy suggested.

Richard

Viorel Popescu

unread,
Jul 12, 2011, 3:22:19 PM7/12/11
to unma...@googlegroups.com
Hi Manuel,

I had the same issue, and I think it stems from the way the covariates are standardized in R vs Presence. If you use 'scale' in R, it yields different values than if you use 'Standardize covariate' in Presence (the 'Scale' option in Presence doesn't seem to do anything). Might also have something to do with the way you setup the Design Matrix in Presence?

Cheers,
Viorel

Manuel Spínola

unread,
Jul 24, 2011, 7:40:40 AM7/24/11
to unma...@googlegroups.com
Thank you very much Richard.

No, I did not standardize covariates in R and PRESENCE.  Is PRESENCE using K?

Best,

Manuel 

2011/7/12 Richard Chandler <richard....@gmail.com>

Manuel Spínola

unread,
Jul 24, 2011, 7:41:57 AM7/24/11
to unma...@googlegroups.com
Thank you very much Viorel.

Best,

Manuel

2011/7/12 Viorel Popescu <vior...@gmail.com>
Reply all
Reply to author
Forward
0 new messages