Starting values in the unmarked frame

[Manuel Spínola]

Dear list members,

I know this topic appeared before but is still unclear for me.

When estimating occupancy in unmarked:

How to use starting values (argument: "starts"?

In which scale they need to be specified?

Is there any example on this?

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]

hi Manuel,
here's an example that I took from the ?occu page -- I just added the starts=c(1,1,1) argument. The starting values need to be on the scale of the parameters as defined in the likelihood and the algorithms perform better if the starting values are close to the MLE. One might be able to pick reasonbale values by looking at the data or taking an educated guess....
regards
andy

> (fm <- occu(~ obsvar1 ~ 1, pferUMF,starts=c(1,1,1)))

Call:
occu(formula = ~obsvar1 ~ 1, data = pferUMF, starts = c(1, 1,
1))
Occupancy:
Estimate SE z P(>|z|)
8.28 23.0 0.360 0.719
Detection:
Estimate SE z P(>|z|)
(Intercept) -1.929 0.164 -11.730 8.92e-32
obsvar1 0.102 0.182 0.563 5.73e-01
AIC: 263.0144

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

To: unma...@googlegroups.com
From: Manuel Spínola <mspin...@gmail.com>
Sent by: unma...@googlegroups.com
Date: 07/02/2011 08:29AM
Subject: [unmarked] Starting values in the unmarked frame

[Manuel Spínola]

Thank you very much Andy.

This means previous to any backtransformation?

The 3 ones belong to Occupancy, Intercept and obsvar1?

Best,

Manuel

2011/7/3 Andy Royle <aro...@usgs.gov>

[Richard Chandler]

Hi Manuel,

When there are covariates in the model, it does not make sense to back-transform the parameter estimates, which will in general be an intercept and some coefficients. You would only back-transform the estimates of "psi" for some combination of covariates. This is what linearComb() and predict() do. For occupancy models, the interpretation of the coefficients is the same as in logistic regression. Suppose you fit the model

logit(psi_i) = beta0 + beta1*x

where x is the covariate, beta0 is the intercept, and beta1 is the coefficient (slope parameter). Let's say you get estimates beta0=-1 and beta1=1. To visualize this in R, you could do this:

curve(plogis(-1 + 1*x), from=-3, to=3, ylab=expression(psi))  # plogis is inverse-logit

The starting values are just some values for the coefficients of the model. The order is psi then p for occu. That should be better documented... and is confusing since the formula is p then psi.

Richard

_____________________________________
Richard Chandler, post-doc
USGS Patuxent Wildlife Research Center
301-497-5696

From:	Manuel Spínola <mspin...@gmail.com>
To:	unma...@googlegroups.com
Date:	07/07/2011 08:41 AM
Subject:	Re: [unmarked] Starting values in the unmarked frame
Sent by:	unma...@googlegroups.com

---

[Manuel Spínola]

Thank you very much Richard,

So, ones (1s) will be good starting values for all the betas in my model?

Best,

Manuel

2011/7/7 Richard Chandler <rcha...@usgs.gov>

[Andy Royle]

hi Manuel,

 as a general rule if you use starting values  of 0 things should work out ok. Otherwise you can guess whether covariate effects should be positive or negative and start them above or below zero (+1, -1, or higher or lower depending on what you think). But 0 is a good place to start.

regards

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: 07/09/2011 10:29AM

[Manuel Spínola]

Thank you very much Andy.

Best,

Manuel

2011/7/9 Andy Royle <aro...@usgs.gov>