Northern Prairie Wildlife Research Center
Careful observations and notes made over several days before and immediately during the capture operation on each pond were used to locate broods and determine brood age (Gollop and Marshall 1954), size, and species composition. Usually only a single brood occurred on a pond; where several broods occurred on a pond, they usually could be distinguished by their different ages. In a number of cases, the brood hen was marked and the hatch date of the brood was known. Ducklings were caught in drive traps (Cowan and Hatter 1952, Trauger 1971) in late July and August. Ducklings were weighed immediately after capture, as soon as the down was dry. Mass was determined to the nearest gram using Pesola spring scales, the accuracy of which was regularly checked with known weights. Age of each duckling was determined in hand using plumage classes described by Gollop and Marshall (1954). Most captured ducklings were 20-50 days old (Classes IC-IIC). Some ducklings were classified only as Class III; because we were uncertain if they were Class IIIA or later, we separately delineated them as Class IIID. Statistical comparisons of body mass among and within broods were made only with data from broods which would be readily identified from others on the same pond. Within each identified brood, we determined mean and range of mass differences by comparing each possible pair of ducklings within the brood.
We used analysis of variance (ANOVA) to assess the effects of year, age class, sex, and season on mean duckling mass for Canvasbacks. Season refers to early- or late-hatching ducklings, where early-hatched ducklings are those hatched from nests initiated during the first 15 days of known Canvasback nest initiation each year (Serie et al. 1992). The design was a split-plot; broods within year-by-season-by-age class combinations were the whole plots, sex was the subplot level, and individual ducklings were the subsampling units. Mean separations following significant effects (P ≤ 0.05) were done using Fisher's protected LSD values at the 0.05 level of significance (Milliken and Johnson 1984).
We assessed factors affecting variances of mass within Canvasback broods using a means model approach (Milliken and Johnson 1984) in an ANOVA. We first analyzed all testable hypotheses for yearly, seasonal, age class, and sex differences simultaneously in an unbalanced splitplot design. Broods served as the whole plot and sex as the subplot. We then tested individual hypotheses (contrasts) using single degree of freedom F-tests.
We used the following ANOVA model to estimate variance components attributable to within versus among broods and to compare between pure and mixed-species broods (brood type) for each species:
Yijk = µi + bj(i) + dk(j(i)) ,
Yijk = weight of duckling k in brood j of type i
µi = mean mass of ducklings of type i,
bj(i) = random variable for brood j of type i, and
dk(j(i)) = random variable for duckling k in brood j of type i
The variance of Yijk = vb² + vd², where vb² is the variance among broods and vd² is the variance within broods. Unbiased estimators of mean mass of ducklings in pure and mixed broods are given as least squares means (LSMEANS). All statistical tests were conducted using SAS (SAS Institute, Inc. 1989).