Доверительные интервалы метода взвешенных наименьших квадратов и стратегия градуировки

We propose to use the calibration strategy based on the method of weighted least squares and a priori information about the measurement system and method of analysis. The strategy allows for the lowest possible detection limits and determinations, as well as for proper assessing of the prediction error of the analyte content from the magnitude of the analytical signal. A priori information including the dependence of the measurement variance on the magnitude of the analytical signal is obtained at stage of validation of the analytical procedure. Method of least squares is used most often to approximate the dependencies in analytical chemistry, photography, physics, economics, and the expected measurement errors are explicitly or implicitly assumed to be the same for all the points. Allowance for the inequality of measurement errors (heteroscedasticity) leads to a variant of the method of weighted least squares, however, the theory of developing the confidence intervals in this case has not yet found wide practical application. To implement the calibration strategy we develop a model program capable of calculating the profile of the dispersion curve and providing plotting of the calibration curves with allowance for heteroskedasticity of the errors.