.
We used a length-frequency distribution of the sexes combined for some years (see the Data section) with appropriate modifications to the above equations.
The log likelihood for the tagging component of the model is:
,
where ~ is again used to denote the observed data.
The final objective function is then:
We programmed this model using AD Model Builder (Fournier 1996). AD Model builder uses the C++ auto-differentiation library for rapid fitting of complex non-linear models, has Bayesian and profile likelihood capabilities, and is designed specifically for fitting these types of models.
.
,
