Use of total body electrical conductivity to predict northern bobwhite lipid mass

Several studies have developed models to predict lipid mass of live birds from a measure of total body electrical conductivity (TOBEC), yet few studies have assessed the validity of these models with independent data. We used independent datasets from Illinois (n = 38), Indiana (n = 94), and Missouri (n = 41) to develop and validate multivariable regression models (calibration equations) predicting lipid mass of live northern bobwhites (Colinus virginianus) from measurements of body mass and TOBEC. We assessed the validity of each model by determining the mean difference between predicted and observed lipid mass for bobwhites used in constructing the other 2 models (prediction error). First-order models (i.e., lacking interactions and quadratic terms) were selected as most useful for predicting lipid mass for all 3 datasets (R-2 = 0.71-0.92, Ps < 0.01). Mean predicted lipid mass was less than or equal to 19.1% (2.32 g) of observed means. Large prediction errors associated with 1 dataset may have resulted from differences in methods. Models for the other 2 datasets (Indiana, Missouri) were relatively accurate: mean prediction errors were less than or equal to 3.4% of observed means. Removing TOBEC from Indiana and Missouri models changed their accuracy little; mean prediction errors for these cross-validated models were less than or equal to 2.3% of observed means. Consequently, the benefits of using models with TOBEC cannot always be assured.