Catch was identified, weighed, and measured. Operational and biological data were collected by DMF biologists, including: catch composition and weights for all species, midline lengths for Atlantic cod (and other species as practical) to the nearest cm, set and haul times, locations, weather conditions, depth, and bottom seawater temperature. Data were entered into a customized Access database and analyzed using the open-source statistical program R (R Development Core Team, 2009; Sarkar 2009).
Holst and Revill (2009) described an implementation of Generalised Linear Mixed Models (GLMM) to paired catch experiments. This implementation allows fitting of curves of limited complexity to expected proportions-at-length (in our case, count of cod in NO pots/total count in both pots for each pair). GLMMs in the Holst and Revill (2009) method incorporate between-pair variance (Fryer, 1991). Four fixed-effect models (constant, linear, 2nd order, and 3rd order polynomial relationships of length) were tested, each using pair as a random effect. We used the penalised quasi likelihood function (glmm-PQL function in MASS package of the R statistical software (R Development Core Team, 2009)), where insignificant terms are removed based on the Wald’s test (Holst and Revill, 2009).
Collected video was reviewed at least twice, by two separate reviewers. Observations of fish were noted, and identified to likely species where possible. Actions of fish relative to entry to the camera frame, direction, level of activity, and activity were noted. Pot motion, direction of current, and visibility of the pot were also recorded.