Miðbiksmat í vistfræðilíkönum -Paul Frater

Titill verkefnis: Interactions among data, process, and parameters in stock assessment models

Doktorsefni: Paul Frater

Ágrip
Non-linear modeling of fish stocks involves many different and varying processes and data types. Data can often be biased or correlated due to the process and method of sampling. Some of the various aspects of these interactions are explored and solutions to some of them are sought. This involves exploring aspects of both growth and selectivity that bias length-at-age data, then a method is presented to reverse this bias by inverting selectivity out of the likelihood estimator. Further, results from a Bayesian model of growth are shown for Icelandic cod across time, illustrating how to proceed when the above method is not available. Finally, simulation studies are discussed, extending some of these issues into stock assessment models. These will include assessing the importance of age data under different scenarios of variance in length-at-age as well as a test to assess likelihood methods for dealing with correlated length data.

Doktorsnefnd:

Gunnar Stefánsson, prófessor við Raunvísindadeild
Bjarki Þór Elvarsson, tölfræðingur hjá Hafrannsóknastofnun, rannsókna- og ráðgjafarstofnun hafs og vatna.
Richard Methot, ráðgjafi við stofnstærðarmat hjá NOAA: National Oceanic & Atmospheric Administration