Balancing moult data by subsampling non-moulting birds prior to regression analysis

Abstract

During the analysis of moult records from the SAFRING database it was found that for some datasets the records were not evenly distributed temporally and the proportion of moulting to non-moulting birds was not what would be expected from random sampling. In an attempt to balance these data, the records of non-moulting birds were subsampled with different sample sizes prior to moult regression analysis, and the resulting moult estimates were then compared. The results suggest that subsampling non-moulting birds such that they occur in the expected proportion to actively moulting birds, based on the duration of moult, provides the best estimates of moult.