Estimating age of young birds with a multivariate measure of body size

ABSTRACr.-We analyzed the relationship between age and body size of 65 Great Blackbacked Gull chicks (Larus marinus) between 0 and 56 days of age (n = 209 sets of measurements) using birds from southern New Brunswick in 1989. The logistic model provided the best fit to the body size and age data; it was used to make linear transformations of body-size variables (culmen, tarsus, and wing cord). The first principal component (PC1) from the correlation matrix of transformed morphological variables also was used as a measure of body size. We estimated linear-regression models of univariate body-size measures and of PC1 versus age, and a multiple-linear-regression model of all body size measures versus age. Accuracy of the predictors was evaluated from a verification sample of chick measurements that were, initially, withheld from the analysis. All predictors had negligible bias in age estimation, but PC1 and multiple-linear-regression models estimated age more precisely than did univariate predictors. The increased accuracy of the multivariate models resulted from the reduced effect of outliers on age estimation. Received 16 October 1990, accepted 13 January 1992.