[Developers] Discrete parameters

[Mark Maunder]

ADMB does not have discrete parameters (e.g. integers), but can some ingenious code be created within a tpl to have discrete parameters when running MCMC (or in the source code)?

For MLE estimation
For a single discrete variable and small number of categories, the model could be run for N phases, where each phase sets the discrete parameter to a category in the BETWEEN_PHASES_SECTION, the likelihood inserted in a vector at the end of the phase (e.g. in the REPORT_SECTION), after the final phase, the vector of likelihoods is searched to fine the category that maximizes the MLE, then one more phase is run that fixes the discrete parameter at that category and re-estimates the parameters. This is a bit like a likelihood profile calculation, so efficiency ideas from that algorithm might help. The standard deviations will be biased low. So perhaps vectors need to be kept for the parameter values for each phase and then the standard deviations adjusted appropriately.

Mark

Mark Maunder
Head of the Stock Assessment Program
Inter-American  Tropical Tuna Commission

President
ADMB Foundation

8604 La Jolla Shores Drive
La Jolla, CA, 92037-1508, USA

Tel: (858) 546-7027
Fax: (858) 546-7133
mmaunder at iattc.org<mailto:mmaunder at iattc.org>
http://www.fisheriesstockassessment.com/TikiWiki/tiki-index.php?page=Mark+Maunder

Visit the AD Model Builder project at
 http://admb-project.org/

See the following website for information on fisheries stock assessment http://www.fisheriesstockassessment.com/


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.admb-project.org/pipermail/developers/attachments/20130124/3e274c00/attachment.html>