AN ASSESSMENT OF DATA POOLING AND SOME ALTERNATIVES

Data pooling is an analytic procedure in which multiple samples of an individual's behaviour are treated as independent events. Although common in animal behaviour research, data pooling has been discredited because it may violate statistical assumptions. Four data sets were analysed in both their pooled and unpooled forms. Pooling did not bias results provided that either intra-subject variance exceeded between-subject variance or Ns were equal. Between-groups tests of significance were affected in the same way as descriptive statistics, and as long as intra-subject variance exceeded between-subject variance, pooling did not increase the probability of a type I error. Utmost care must be taken to sample individuals from populations and behaviour from those individuals in an unbiased manner.