The aim of this study was to assess the diierences in the values of the pharmacokinetic parameters attributable to the use of either linear or nonlinear regression analysis and to ¢nd the eiect of weighting schemes on these diierences. Six calves received 20 mg/kg oxytetracycline i.v. Blood samples were drawn during 72 h. The assay of the drug was performed microbiologically. A bicompartmental pharmacokinetic model was used, kinetic analysis being carried out by linear regression (LR) and by weighted least-squares nonlinear regression (WLSNLR). Statistical analysis included a test for normality, the Kruskall^Wallis test and ANOVA with log transformation. The A0,a and B0 did not show any statistically significant differences attributable to the mathematical method used. On the other hand, the statistically signi¢cant diierences in the b values found using the Kruskall-Wallis test and ANOVA with log transformation could be attributed to the diierent methods employed. ANOVA with log transformation determined statistically signi¢cant diierences between the parameters obtained by linear analysis and those obtained by WLSNLR when the weighting (w) was 1. When weights were 1/x ,1 / x 2 or 1/Hx, no statistically signi¢cant diierences were found. The optimal weighting scheme was w=1/x 2 because of a more homogeneous scatter and random distribution of residuals aboutthe abcissa axis in a plot of weighted residuals in the ordinate versus time in the abcissa.It was concluded that the use of these different procedures can give major variations in the apparent value of b, the most important pharmacokinetic parameter. The correct selection of the weighting procedure is therefore fundamental in obtaining the best estimate of this pharmacokinetic parameter inWLSNLR.