ESTIMATING SEX-RATIOS WITH DISCRIMINANT FUNCTION-ANALYSIS - THE INFLUENCE OF PROBABILITY CUTPOINTS AND SAMPLE-SIZE

The influence of different probability cutpoints and associated variations in sample size on population sex ratios were inferred with birds sexed by discriminant function analysis (DFA). Using two data sets from Dunlins (Calidris alpina) collected in western Washington state as examples, it was observed that selection of high probability (greater-than-or-equal-to 0.95) cutpoints resulted in biased population sex ratio estimates. Overall, approximately 80% of the samples from the two data sets (in the probability cutpoint range of greater-than-or-equal-to 0.7) used in the analyses were required for accurate estimation of sex ratios. Calculation of sample sizes required for a 5% error bound on the estimate of the proportion of females in the study population indicated that 200-300 samples were required. This figure will probably vary for other species of birds with different sex ratios and population sizes. A technique from the statistical literature for assessing the statistical limitations of sex ratios based on data sets with fixed sample sizes is presented. DFA has great potential as a passive, physically nondestructive technique for sexing species of monochromatic birds, but sample size must be considered when population parameters such as sex ratio are inferred using birds sexed by DFA.