Application of logistic regression analysis to predict cannibalism in orange-spotted grouperEpinephelus coioides(Hamilton) fry

Abstract

Cannibalism is frequently observed in larviculture of orange‐spotted grouper Epinephelus coioides. Previously, based on measurements of morphometric characters, a linear equation of total length (TL) of prey to cannibals was proposed: TLₚᵣₑy = 0.80 TLcₐₙₙᵢbₐₗ – 1.50. To verify the reliability of the equation, experiments were performed with pairs of fish with different TLs. Cannibalism occurred only when the cannibal‐prey size ratios were equal to or larger than that predicted by the equation. To predict the probability of cannibalism among the grouper of known TLs, a logistic regression model was proposed. The logistic regression model is: [Formula: see text] The logistic regression model showed the following: when either TLₚᵣₑyor TLcₐₙₙᵢbₐₗis constant, the probability of cannibalism increases with increase in the cannibal‐prey size ratios; if given a constant cannibal‐prey size ratio, probability of cannibalism is lower in early stages than in later stages. The prediction contrasts with that calculated from the linear equation where the minimum cannibal‐prey size ratios decrease with size of the cannibal. However, the prediction matches observed rearing experiences: cannibals prefer smaller prey to larger ones and the cannibalism rate declines as fry age.