Premium
Failure of Statistical Tests to Detect Assumption Violations in the Mark–Recapture Population Estimation of Brook Trout in Adirondack Ponds
Author(s) -
Cone R. Scott,
Robson Douglas S.,
Krueger Charles C.
Publication year - 1988
Publication title -
north american journal of fisheries management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 72
eISSN - 1548-8675
pISSN - 0275-5947
DOI - 10.1577/1548-8675(1988)008<0489:fosttd>2.3.co;2
Subject(s) - salvelinus , mark and recapture , trout , homogeneous , population , statistics , fontinalis , fish <actinopterygii> , fishery , sampling (signal processing) , estimation , biology , ecology , mathematics , geography , demography , computer science , engineering , systems engineering , combinatorics , sociology , filter (signal processing) , computer vision
Misleading mark–recapture population estimates were obtained in three field comparisons of the performance of the Assinica and Temiscamie strains of brook trout Salvelinus fontinalis . Three and possibly four of the six assumptions required for the Jolly–Seber method of population estimation were violated. First, violation of the assumption that marked and unmarked fish have homogeneous probability of being captured was detected by comparisons of the proportion of marked trout present in seining samples with the proportion present when the ponds were drained. Second, violation of the assumption of homogeneous probability of recapture was detected by the Leslie test. Third, violation of the assumptions of homogeneous survival both between marked and unmarked fish and among marked fish occurred and was probably the primary cause of the misleading estimates. Petersen estimates based on recapture of marked fish at pond draining were calculated for comparative purposes and also failed to accurately gauge the size of these brook trout populations. Inaccurate population estimates, deviating by as much as 41% from the actual numbers, might have been accepted if the study had relied solely on data from sampling for the statistical tests of the assumptions. Assumptions were violated even though our sampling covered 45–75% of the pond areas and samples contained substantial portions (20–30%) of the total population. Careful validation of the assumptions of mark‐recapture models is stressed as prerequisite to the acceptance of numerical estimates of wild populations.