Analytical methods require adequate curve adjusting models to express reliability. The ordinary or weighted least squares methods (OLSM or WLSM, respectively) were used to determine the most adequate mathematical model for the analytical curve, beginning from the simplest method (linear) to quadratic. These models were evaluated with respect to normality and homoscedasticity of the residues. Analytical curves were built by injection of 1, 5, 10, 15 and 20 mL of simvastatin at 40 µg mL-1 (40, 200, 400, 600 and 800 ng) or of 10 mL of simvastatin at 4, 20, 40, 60 and 80 µg mL-1, employing high-performance liquid chromatography with photo-diode array detection (l 238 nm). The best-adjusted models were the linear and the quadratic, when observed in terms of the dataset masses and concentrations, respectively. In the considered range, WLSM was more appropriate than OLSM. The different behavior indicated the need for careful selection of the most adequate model to express the analytical curve and the need for assuring the reliability of the used method.
analytical curves; HPLC-UV; least squares method; weighted models; validation
