Statistical model for estimating measurement error in the congener-specific analysis of PCBs by gas chromatography

In this study, we use repeated measurement to estimate the measurement error of congener-specific analysis of polychlorinated biphenyls (PCBs) by gas chromatography according to a proportional measurement error model. The estimator of measurement error performs extremely well when the measurement is made without systematic error. The estimator also smoothes the distribution of measurement error. The Q-Q plots illustrate that the error term, according to the model, follows a stable law of distribution which is approximately log-normal and has light tails at both ends due to larger deviation. The parameters of the distribution and tails vary with different GC devices and different kinds of samples as well. Given the shape of the Q-Q plot of measurement errors, the tolerance interval of the PCB measurement can be found without the normally distributed assumption which is necessary in calculating the traditional confidence intervals.