Estimation of Absolute and Relative Survival Rates from the Recoveries of Dead Animals
- Biometrics Unit, University of Otago, P.O. Box 56, Dunedin, New Zealand
Circumstances that give rise to samples of dead animals from natural populations are considered and five important particular situations are emphasized. In two of these situations it is possible to estimate the absolute mortality rates of animals in the natural populations concerned. In the other three situations the populations comprise two or more different types of animal and only the relative mortality rates of these can be estimated.
The most obvious examples of the first two situations come from bird banding experiments. Models for such experiments are therefore briefly reviewed. A Poisson model for samples of dead animals from a population with an unknown initial size is proposed and is shown to produce survival rate estimates that can be readily calculated on the assumption that the survival rate per unit time becomes constant for older animals. This model is of value since the estimation does not require iterative computer calculations whereas other models making essentially the same assumptions do require these.
The third and fourth situations that have been considered concern large populations with relatively small numbers of deaths. The relative mortality rates of the different types of animals in the populations can be estimated by comparing the proportions of dead animals of the different types with the corresponding proportions of live animals.
The final situation discussed occurs when animals have associated with them values for certain characters X1, X2, . . ., Xp and the relative mortality rates of animals with different X values is to be determined by comparing the distribution of the X's for live and dead animals.