The Effects of Trap Response on Tag Recapture Estimates
In recent years capture-tag-recapture experiments have been widely used for estimating the size of animal populations (cf. Cormack  for an excellent review). However, such methods are only applicable when a number of restrictive assumptions are known to be true, or at least approximately true. For example, it is generally assumed that all animals are equicatchable, and that trapping and tagging do not affect future catchability. However, in practice catchability may vary from individual to individual, e.g. variations between sexes, age groups, species (Kikkawa ); gear selectivity in fish populations (International Commission ); the effect of bait, trap, and habitat preferences upon catchability (Chitty and Shorten , Corbet ); and the effect of trap spacing on catchability (Kikkawa ). Also trapping itself can affect the future catchability of animals as they may become 'trap-shy' or 'trap-addicted,' e.g. Evans , Geis , Morris , Flyger , Huber , to mention just a few. Trap respongse can often be minimised by using prebaited traps (Chitty and Kempson ), though this does not always work (Young et al. , Croweroft and Jeff ers ). Therefore in spite of careful experimentation the experimenter can never be sure that the underlying assumptions necessary for statistical analysis are satisfied. It is then imperative that the experimenter know something about the robustness of his statistical methods with respect to departures from the assumptions and where possible to test for these departures. As a step in this direction we shall examine the effect of certain departures on the so called Petersen estimate (commonly called the Lincoln index, cf. Le Cren ) and discuss briefly a regression method of Marten  for detecting these departures.