Fig. 1. The overall bill length material described in the text above; n = 180 (96 males, 84 females). Males white columns, females black, from base line; mean values shown with arrows. [CP]
The material presented here consists (16.2.03) of bill-length measurements from 66 (42 males, 24 females) migrant Dunlin collected at Gdansk 25.7 - 29.9.91 and sexed by dissection (J. Gromadzka), and 104 breeding birds (52 males, 52 females) from the Yaibari area, Yamal peninsula, W. Siberia (measurements by H. Behmann and M. Gromadzki). In the latter case birds were sexed by means of morphological characters, cf. Ferns & Green 1979 and Stiefel & Scheufler 1989, I use them here because I am convinced that the accuracy of determination is close to 100 %. These two materials were combined with the bill-lengths of 10 genetically/morphologically sexed individuals in Rösner 1997 and Wenink 1994, altogether creating a sexed reference material of 180 birds (96 males, 84 females). I will continue this collecting of material till I have an adequate reference population for Baltic/Waddensea conditions, or maybe better: one for the Baltic and one for the Waddensea. I will take anything that was sexed by genetical means or dissection, and credit will be given to the sources.
I. For a test discriminant functions were calculated from Behmann's Yamal material (the only one with weights, wing and bill measurements): 30 males and 29 females. The sex in 27 out of 30 males (90 %) and 25 out of 29 females (86.2 %) was correctly predicted with the discriminant functions (male) 1.78 WEIGHT + 2.37 BILL + 11.86 WING - 781.31 and female 1.88 WEIGHT + 3.54 BILL + 11.92 WING - 831.23 (the bird belongs to the sex that gives the highest value of either function, I call this a TYPE A discriminant function). If multiple regression is calculated from the same material with male coded +1, female -1 and weight, wing and bill as predictors, the regression equation will be DS = 11.7 - 0.013 WING - 0.024 WEIGHT - 0.274 BILL; general constant and bill constant highly significant, wing and weight constants not significant (null hypothesis that they are 0 can not be rejected), r2(adj) = 52.2 %. (TYPE B discriminant function) Birds with DS > 0 should be males (scores 0.13 to 1.57 and -0.07, -0.14, -0.67), DS < 0 females (scores -1.35 to -0.05 and 0.04, 0.36, 0.47, 0.59); again 3 males and 4 females are not correctly identified. The corresponding equation calculated by Brennan et al. 1984 for wintering C. a. pacifica in Washington was DS = 34.5052 - 0.0893 WING - 0.0653 WEIGHT - 0.5381 BILL, here 91.5 % of 200 birds are correctly identified, against 88.1 % in the Yamal material. In most cases it seems as if 9 out of 10 Dunlin can be correctly sexed with discriminant functions, and at least weight can be omitted as predictor in the W. Palearctic, it only confuses the calculations.
II. The complete Yamal material (52 males, 52 females) with sexes coded 1 and -1, wing and bill as predictors, has the discriminant function DS = 14.6 - 0.046 WING - 0.279 BILL, p < 0.0005 for general constant and bill constant, p = 0.056 for wing constant, r2(adj) = 56.5 %. 86.5 % of the females and 88.5 % of the males are correctly identified by the equation.
III. From Gdansk 5.8 - 29.9.91 there are 28 1c males and 20 1c females with wing/bill biometry, sexed by means of dissection. With sexes coded 1 and -1, wing and bill as predictors, the discriminant function is DS = 21.5 - 0.110 WING - 0.256 BILL, p < 0.0005 for general constant and bill constant, p = 0.01 for wing constant, r2(adj) = 47.2 %. 80 % of the females and 89.3 % of the males are correctly identified by the equation.
It should be added, that males are sexed with 95 % confidence below bill-length 30.5 - 31 mm, females with 95 % confidence above c34.5 mm, with an average alpina material. In between these limits the probability for correct sexing is less - unless wing lengths help up. The advantage of the discriminant function approach is that confidence values may be calculated in each single case. [CP]