SITE PARAMETERS' VARIANCE - A MISSING DIMENSION IN THE STUDIES OF NEST-SITE SELECTION?

In most studies of nest-site selection the data of habitat parameters are treated with analysis of variance. A basic assumption of this test is the homogeneity of variance. Here, we show that the nest-site selection process leads to lower variance of the selected parameters than in the case of random points which generally describe the available average characteristics of the environment. Thus, the variance should be accounted for in studies on nest-site selection and it should be treated not as a problem (as it is usually done), but as a source of additional important information on the selection process. Comparison only of mean values often does not lead to significant differences between nest site parameters and random points which may result from a small effect size (when animals select features similar to the general mean of available characteristics). Deeper insight into variance of the site parameters may elicit important results. We illustrate this issue with real data on nest site (islets and shores of water reservoirs) selection in the Common Gull Larus canus. Four (islet's area, vegetation height on islets, vegetation cover on shore and distance to nearest shrub or tree on shore) from eight parameters were favored by the birds and, as predicted, their variance values were lower than of those not selected (vegetation cover on islets, distance of the islets to shoreline, vegetation height on shore and distance to water).