Laserfiche WebLink
The covariate TL (length at first capture) was standardized with a z transformation, i.e., <br />standardized length (SL, but not the measure of fish standard length) for fish i was <br />SL; _ (L; - L...)/SD, <br />where L; was the TL of an individual fish, L.. was the mean TL for all fish in the population, <br />and SD the population standard deviation of TL. The z transformation was used because the <br />numerically small covariates that result allows for better model fitting. Because adult Colorado <br />pikeminnow grow very slowly (Osmundson et al. 1997), use of length at first capture as the <br />covariate was deemed appropriate for fish that may be captured several times. Inclusion of the <br />covariate TL in abundance or survival estimation modeling was important because of the <br />potential effects of fish size on probabilities of capture. Abundance estimators such as those in <br />program CAPTURE (White et al. 1982) do not have the capability to use individual covariates <br />because the likelihood includes probabilities for animals that are never captured, so the <br />covariates are unknown. Selection between models was performed with information-theoretic <br />procedures (Akaike's Information Criterion adjusted for small sample size (AICJ, Burnham and <br />Anderson 1998). Analyses demonstrated that capture probabilities were equal to recapture <br />probabilities among the short-term and annual sampling occasions (i.e.,pk = co, so heterogeneity <br />due to capture effects was assumed minimal. We also tested for differences in rates of capture <br />and recapture in various models to evaluate if behavior effects (e.g., fish avoidance of boats) <br />were influencing recapture rates. We initially also fit mixture models of Pledger (2000), which <br />were designed to incorporate heterogeneity caused by differing probabilities of capture for <br />different segments of the population. We abandoned these models because only higher numbers <br />22