2.5. Statistical analysis
All models were obtained by fitting the data to Eqs. (1)–(3) with the least squares and Marquardt algorithm for linear and nonlinear models, respectively. Confidence intervals were calculated by using the Student’s t-value. Residual analysis and lackof fit test were used to assess the statistical acceptability of the models. Kolomogorov–Smirnov test was performed to verify the normality of the residuals, Cochran test for variance homogeneity, and Durbin–Watson test for autocorrelation. All the tests were performed by using STATGRAPHICS
s
Plus version 4.0 (Statistical Graphics Corp).