Flexibility in Survival Analysis from Recapture Data: Handling Trap-Dependence
Résumé
It is well known that survivorship and behaviour of animals can be affected by handling. The deleterious effect of capture can be minimized by taking special care during handling, but it is in general more difficult to avoid changes in behaviour; "trap-shy" animals will tend to avoid the traps while "trap-happy" animals will seek them, particularly when they are baited. General trap-dependence can be treated as age-dependence provided the data are suitably set out. However, trap-dependence effects are likely to be strongest immediately after a capture event. Thus models that consider a different capture probability whether an animal was caught on the previous occasion are particularly relevant. If we d~note by m an immediate trap-effect (following Sandland and Kirkwood, 1981) the straightforward generalization of the Cormack-Jolly~Seber model (St,Pt) can be denoted as (St,Pm*V· I present a goodness-of-fit test for this model and a test of comparison with the Cormack-Jolly-Seber model as simple tests on 2*2 contingency tables. The model (St,Pm*V which allows different trap-effects in different years, such as trap-shyness alternating with trap-happiness cannot be used without caution to provide reliable parameter estimates due to identifiability problems. But it is most often sufficient to consider a model with an additive time and trap-effect on a convenient scale (St,Pm+V· The models (St,Pm•t>, (St,Pm + t) and (St,Pm) can be useful not only when genuine trap-dependence is involved but also to account for other structural effects that mimic trap-dependence or can be alleviated through it, in particular general heterogeneity of capture probabilities among individuals.