of Different Implementations of Catch at Age Models
purpose of this page is to serve as a shared repository for ADModel
Builder code used for analyses of fishery data.
models are intended to provide example applications only. They do not
represent complete models used for stock assessment and fishery
evaluations by any management body.
- A Ricker
stock-recruitment curve fit with explicit errors in stock size. This
is a very simple example of how errors-in-variables type problems
can be addressed. Click here for the template and datafiles.
- Simple catch-at-age
using flack lake trout data (from Hilborn and Walters). This model
is one example that ships with ADModel
Builder. A zipped up version can be found here.
- An age structured
model for Gulf of Alaska sablefish. This example model is of the
aggregate population (sexes combined) but is tuned to sex-specific
length frequency data. This has examples of ways to bin different
age groups and demonstrates how modeled age classes can be processed
to fit observed length frequency data. The template
and datafiles can be viewed
and Richards (1995) CJAFS model is implemented in this example applied
to Pacific ocean perch. Here, the model formulations drawn up are
directly from their paper but the objective function was modified
slightly to avoid problems age classes that had no observations. Note
the use of "posfun()" to avoid having the stock attain
negative values during the non-linear search routine. The template,
pin(parameter input file), and the data
files. These files were kindly provided by Laura Richards, DFO,
Canada. An alternative form for this model (using Baranov's catch
equation can be seen here
with data file here.
- Age structured model for eastern Bering Sea yellowfin sole. This
application (template file) is
tuned to a single survey (for which catchability is estimated). The
organization of the datafile
follows the convention of using calendar years for valid indices of
observation arrays. This makes using data with missing years easier.
- Age structured model
used for evaluating the NRC's (National Research Council's) Expert
Panel on Fisheries Stock Assessment datasets. This model represents
a fairly comprehensive treatment of fishery and survey data where a
wide variety of processes are modeled. For example, survey AND
fishery catchability are allowed to change over time and effort was
assumed to be measured with error. During the last phases of
minimization a robust objective function. The write-up of this model
can be found in zipped-postscript (260k) form here
The template file can be viewed here
with one of the NRC data files shown here.
This code makes extensive use of what I have called control flags.
These serve a variety of purposes from specifying prior variances to
certain processes to performing retrospective analyses.