Frequently Asked Questions, And Others

Description

A place for hints and miscellaneous advice.

How do I install secr?

Under Windows, it is simplest to install the package binary from the Rgui. Save the file secr_1.3.x.zip to a local folder and use the menu option "Packages | Install packages(s) from local zip files...".

For other systems (see below) you will need to install the source package secr_1.3.x.tar.gz.

Whatever your system, you also need to get the package abind (use Packages | Install package(s)... to download from CRAN). Other required packages (MASS, nlme, stats) should be available as part of your R installation.

Does secr work on operating systems other than Windows?

Yes, at least in theory. From version 1.2.8 you should be able to install the source package secr_1.x.x.tar.gz if you have the necessary tools (check the 'R Installation and Administration' manual or the 'R for Mac OS X FAQ' on CRAN for help on installing source packages). Thanks to Ray Brownrigg and Joanne Potts for vital assistance. I would be glad to hear of successful installations on non-Windows systems (particularly if anyone produces a binary for Mac OS).

How can I get help?

There are three general ways of displaying documentation from within R. Firstly, you can bring up help pages for particular functions from the command prompt. For example:

? secr.fit

Secondly, help.search() lets you ask for a list of the help pages on a vague topic. From R version 2.8.0 you can use just ??. For example:

?? 'linear models'

Thirdly, you can display various documents:

RShowDoc ('secr-manual', package='secr')

RShowDoc ('secr-overview', package='secr')

See below for more R tips.

How should I report a problem?

If you get really stuck or find something you think is a bug then please report the problem. For the moment, the correct address is density.software@otago.ac.nz.

When reporting problems please send an actual dataset (ideally, the simplest one that exhibits the problem). Use save to wrap several R objects together in one .RData file, e.g., save('captdata', 'secrdemo.0', 'secrdemo.b', file = 'mydata.RData'). Also, paste the output from packageDescription("secr") into the text of your message.

Needless to say, we cannot promise to solve all problems.

Why do I get different answers from secr and Density?

Strictly speaking, this should not happen if you have specified the same model and likelihood, although you may see a little variation due to the different maximization algorithms. Likelihoods (and estimates) may differ if you use different integration meshes (habitat masks), which can easily happen because the programs differ in how they set up the mesh. If you want to make a precise comparison, save the Density mesh to a file and read it into secr, or vice versa.

Extreme data, especially rare long-distance movements, may be handled differently by the two programs. The 'minprob' component of the 'details' argument of secr.fit sets a threshold of probability for capture histories (smaller values are all set to minprob), whereas Density has no explicit limit. In the current version the default minprob has been reduced from 1e-20 to 1e-50. If you find a discrepancy with Density it may be worth lowering minprob even further.

How can I speed up model fitting and model selection?

If you don't need to model variation in density over space or time then consider maximizing the conditional likelihood in secr.fit (CL = TRUE). This reduces the complexity of the optimization problem, especially where there are several sessions and you want session-specific density estimates (by default, derived returns a separate estimate for each session even if the detection parameters are constant across sessions).

Check the extent and spacing of the habitat mask that you are using. Execution time is roughly proportional to the number of mask points. Default settings can lead to very large masks for detector arrays that are elongated 'north-south' because the number of points in the east-west direction is fixed. Compare results with a much sparser mask (e.g., nx=32 instead of nx=64).

Do you really need to fit that complex model? Chasing down small decrements in AIC is so last-century. Remember that detection parameters are mostly nuisance parameters, and models with big differences in AIC may barely differ in their density estimates. This is a good topic for further research - we seem to need a 'focussed information criterion' (Claeskens and Hjort 2008) to discern the differences that matter.

Use score.test to compare nested models. At each stage this requires only the more simple model to have been fitted in full; further processing is required to obtain a numerical estimate of the gradient of the likelihood surface for the more complex model, but this is much faster than maximizing the likelihood.

Will secr soon include feature X?

Finite mixture models

Possibly. Mixture models do seem desirable as an option to allow for unmodelled heterogeneity. However, mixtures add an extra level of complexity across the entire detection model structure, and an additional 'real' parameter (for the mixing proportion) so the decision is not trivial. A decision was made to release and evaluate secr 1.3 before adding even more complexity.

Inverse prediction

Yes, from version 1.2.7. Fitting SECR models by simulation and inverse prediction is still the only rigorous method for dealing with data from single-catch traps. However, maximizing the likelihood as if the data came from multi-catch traps yields fairly good density estimates even with single-catch data (estimates of detection g0 are biased). Limited capability for simulation-based fitting is provided by the function ip.secr. It is not intended to extend ip.secr to include a wider range of models.

Things You Might Need To Know About R

The function findFn in package sos lets you search CRAN for R functions by matching text in their documentation.

There is now a vast amount of R advice available on the web. For the terminally frustrated, 'R inferno' by Patrick Burns is recommended (www.burns-stat.com/pages/Tutor/R_inferno.pdf). "If you are using R and you think you're in hell, this is a map for you".

Method functions for S3 classes cannot be listed in the usual way by typing the function name at the R prompt because they are 'hidden' in a namespace. Get around this with getAnywhere(). For example:

getAnywhere(print.secr)

R objects have 'attributes' that usually are kept out of sight. Important attributes are 'class' (all objects), 'dim' (matrices and arrays) and 'names' (lists). secr hides quite a lot of useful data as named 'attributes'. Usually you will use summary and extraction methods (traps, covariates, usage etc.) to view and change the attributes of the various classes of object in secr. If you're curious, you can reveal the lot with 'attributes'. For example:

data(captdata)

traps(captdata) ## extraction method for 'traps'

attributes(captdata) ## all attributes

Author(s)

Murray Efford murray.efford@otago.ac.nz

References

Claeskens, G. and Hjort N. L. (2008) Model Selection and Model Averaging. Cambridge: Cambridge University Press.