Influence of method of model fitting and gender on estimated parameters for the relationship between starvation heat production and body size in the domestic fowl.

The relationship between starvation heat production (SHP) in kJ/d and body weight (W) in kg for domestic fowl was examined with compiled calorimetric data for 294 fowls (Johnson and Farrell, 1985). Linear regression analyses using ordinary least squares methods were performed before and after transformation of the data to a logarithmic scale (base = 10). Coefficients of determination (R2) were similar for untransformed and log-transformed data (R2's = .56 and .53, respectively). The slopes for untransformed data of males and females were different (P less than 0.033). When the model parameters for: SHP = b0 Wb1 were derived from log-transformed data, there were no significant differences in exponents for males vs. females (P greater than .95). However, when the coefficients were estimated using non-linear regression techniques, a difference was detected at the level of P = 0.072. It was determined that there is no advantage in describing the relationship of SHP as a non-linear function of W, nor is it appropriate to express physiological variables as a ratio of weight raised to some mass exponent without statistical justification. Using W or W2 as a covariable for the statistical analysis of physiological data would produce equally acceptable determinations of significance.