Ardea
Official journal of the Netherlands Ornithologists' Union

login


[close window] [previous abstract] [next abstract]

Perdeck A.C. (1985) Methods of predicting fat reserves in the Coot. ARDEA 73 (2): 139-146
This study deals with methods to assess the fat reserves in live birds. The basic method used is to relate, in a source group, the fat-free weight to certain body measurements with a multiple regression analysis. With the resulting regression formula fat-free weights were predicted in new individuals, forming a test group. The predicted amount of fat can then be found by subtracting the predicted fat-free weight from the total fresh weight (fat-free weight method). The amount of fat can also be related directly to total fresh weight and body dimensions (fat method). When no absolute amounts of fat are needed, but this amount is to be corrected for individual size, a fat index can be used. It is shown that a fat index can be estimated in various ways, basically derived from several regression equations. Each of these methods has been applied to a sample of Coots, of which 73 belonged to the source group, and 20 to the test group. To compare the predictive precision of the methods the standard error of estimate (better: prediction) of the test group was used. For predicting the absolute amount of fat the fat-free weight and the fat method are of the same predictive power (Table 2), but the latter might be preferred since it corrects for size differences between sexes and age groups, while fat-free weight methods needs knowledge to sex and age of each individual in such situations. With the index methods the standard error of prediction varied from 0.084 to 0.238, but relative to the mean fat index, from 0.57 to 0.72 (Table 3). The figure of 0.57 belongs to the dry index method, where the amount of fat divided by the total dry weight is, as a ratio, directly related to total fresh weight, wing width and tarsus length. This is a rather inaccurate prediction (57 per cent error of mean actual index). Compared to other published studies, however, the amount of variation explained by the regression equations estimating fat-free weight or amount of fat is quite satisfactory. However, the precision of prediction is lowered both by the application to a test group, which is in fact for what the methods are developed. To check the reliability of this final result original data of other studies were obtained. They concern the Great Crested Grebe and the Dunlin. Application of the methods of these data resulted in absolute and relative standard errors of prediction of the same magnitude as found in the Coot (Table 4 and 5, Fig. 3 and 4). There are indications, however, that absolute errors vary with species. They are rather high in Coot and Great Crested Grebe, less in Dunlin, and even smaller in warblers, to judge from few data published (Rogers & Odum 1946). But in general the error made by a prediction in a new individual is in the order of 25 to 70 per cent of the actual fat amount or index, depending mainly on method and magnitude of the fat amount or index. This forces one to apply the methods to groups rather then to separate individuals, unless the differences in fat or fat index are large as compared with the absolute errors. It is advisable to check the precision of the predictions on test individuals that were not included in the original regression models. There is a need to develop more accurate methods.


[close window] [previous abstract] [next abstract]