3.6.4. Calibration Curves

Seven-point calibration curves were constructed in triplicate by plotting the analyte peak areas versus different analyte concentrations of NDMA, NDEA, DIPNA, or EIPNA. The weighted-linear least square model with a 1/x² weighting factor was selected using a statistical test of homoscedasticity.²⁷ The coefficient of determination (r² ≥ 0.995) was determined to verify the linearity of the plot. The percent deviation of the mean back-calculated concentrations of NDMA, NDEA, DIPNA, and EIPNA should be within ±10%. The residual plot was demonstrated whether the slope and y-intercept were significantly different from zero at a 95% confidence interval. Statistically, a p value of the slope should be less than 0.05, meaning that it is significantly different from zero, whereas the p value of the y-intercept should be more than 0.05, meaning that it is insignificant from zero. F_cal should be greater than F_ANOVA demonstrating the linear regression.²⁸