Evaluating methods for estimating process and observation error variances in statistical catch-at-age analysis

Simulation testing the robustness of stock assessment models to error: some results from the ICES strategic initiative on stock assessment methods

“…More specifically, this methodology has been used to examine issues associated with data availability, model misspecification (i.e. structural uncertainty), and the effect of observation and process errors (ICES, 2004;Linton and Bence, 2008;Wetzel and Punt, 2011;Deroba and Schueller, 2013). Much of this previous simulation work, however, was based on generic fish populations or was designed for applications to specific fish stocks (Kell et al, 1999;Haltuch et al, 2008).…”

Revisiting data weighting in fisheries stock assessment models

“…The first departure is the use of what have come to be called state-space models (e.g., Millar and Meyer, 2000;Linton and Bence, 2008;Nielsen and Berg, 2014; also see de Valpine, 2002 for an excellent introduction to the associated theory, but note that this field is rapidly evolving because of substantial technological advances since 2002). In discussing these models I will use the term process variation to designate what is usually called "process error" in the state-space literature, because I have used the latter term in a different sense (see definition above).…”

“…", and, when there are multiple time-varying processes, (iii) "do we have the right balance of weights (i.e., process variances) amongst these processes?". Linton and Bence (2008) used a simulation experiment to show that (ii) was usually difficult to answer in their state-space statistical catch-at-age model (in their terminology we would say that they could not usually produce good estimates of both process and observation process error variances [NB they did obtain good estimates in the case when an informative prior was used for the process error variance, but noted that "it is unlikely that a stock assessment analyst would have the necessary data to set such an informative prior"]). It is well understood that getting (i) wrong can strongly affect the outcome of a stock assessment.…”

From spawner habitat selection to stock‐recruitment: Implications for assessment

“…A variety of factors influencing SR estimates has been suggested, e.g. time-series bias (Walters, 1985), observation (Walters & Ludwig, 1981), and process errors (Linton & Bence, 2008), productivity regimes (Gilbert, 1997;Vert-pre et al, 2013), and nonstationary dynamics (Feiner et al, 2015;Quinn & Deriso, 1999).…”

“…A variety of factors influencing SR estimates has been suggested, e.g. time‐series bias (Walters, 1985), observation (Walters & Ludwig, 1981), and process errors (Linton & Bence, 2008), productivity regimes (Gilbert, 1997; Vert‐pre et al, 2013), and nonstationary dynamics (Feiner et al, 2015; Quinn & Deriso, 1999). Biological and ecological aspects such as age structure, spatial distribution, fecundity and spawning patterns are also known to influence the variation in recruitment (Green, 2008; Shelton et al, 2012).…”