The relative bias in sampling estimates of production and disease parameters of caged Atlantic salmon

The relative selection bias in dip net sampling of caged Atlantic salmon was investigated at two different stages of the production cycle. A pre-grading sample was obtained by crowding a group of fish within their cage and collecting three to five at a time in a dip net.

The mean fish weight estimated by the dip-net sampling method was significantly heavier (102 grams, p = 0.013) than that estimated by the systematic random sampling method at one site. The dip sample mean fish length was significantly shorter (.82 cm, p = 0.0001) at one site and significantly longer (0.21 cm, p = 0.044) at the other site when compared to the random sample. Condition factor was significantly greater (0.099, p = 0.0003) at one site only. Within individual cages, the dip sample estimates of means ranged from 145 grams lighter to 355 grams heavier, from 1.7 cm shorter to 0.7 cm longer, and from 0.04 units less to 0.26 units more for condition factor when compared to estimates from random samples.

Prevalence of granulomatous lesions at necropsy was significantly higher (p $<$.05) in dip samples than in systematic random samples. Lesions were also significantly associated with smaller fish in one cage (p $<$.05). (Abstract shortened by UMI.).