Testing and improving the accuracy of discriminant function tests: a comparison between morphometric and molecular sexing in Macaroni Penguins

Dimorphism in various morphological traits is widely used to sex birds. However, when a threshold is set and there is overlap in the distribution of traits between sexes, there will be an error rate. Macaroni Penguins show limited sexual dimorphism in size, with the distributions of most phenotypic traits overlapping between sexes, making using morphological characters for sexing difficult, although bill depth is used ill one discrimination test. Two existing DNA-based PCR tests that use different primer pairs (P2/P8 and 2550/2781) were adapted for use with fluorescently labelled primers to enable the size of fragments to be accurately measured via genotyping protocols and sex to be determined. The two Molecular tests were applied to the same samples to estimate the error rate within and between tests. Whilst both primer pairs reproducibly identified the sex of the penguins tinder Study With 110 inconsistencies between tests, tire P2/P8 primer pair had a higher Success rate of amplification. To estimate the error rates of morphometric tests, they were compared with molecular tests, revealing areas of disagreement. Structure in the error indicated a bias near the morphometric discrimination threshold. We suggest that larger sample sizes in discriminant function analyses and calibration with molecular data can be used to estimate the likelihood of misidentification for any particular value of the trait. A description of where error lies can then be used to target the more expensive and time-consuming DNA tests to check individuals that are most likely to be mis-assigned. Received 15 September 2008, accepted 9 January 2009