Applied statistics, microcomputers and analytical chemistry

An applied statistics software package, containing a unique weighted linear regression (WLR) routine, has been developed. Its features are demonstrated using real and Monte Carlo simulated data. The WLR routine is particularly useful for absolute and comparative calibrations. In the absolute calibration of analytical systems the statistical analysis of the linear calibration curve produces an analysis of variance (ANOVA), the curve's equation and confidence band, confidence limits for the slope the regression coefficient, and intercept, and the standard error of regression. The routine also computes an unknown's concentration and its "confidence" limits. 80th simple (SLR) and WLR can be used for absolute calibration. WLR can be used without knowing the error's standard deviation (SO); assuming the analytical error is normally distributed, the SO is a linear function of concentration or response, and the concentration range of interest is well above the detection limit. Under these conditions the computed "standard error of regression" is the "relative SO" when using WLR, or the "SO" for SLR. Comparative calibration is used for method validation and for determining the relative economic and technical merits of analytical systems. Ways of estimating a system's precision, as a function of concentration. are discussed. Two new, simple approaches are demonstrated. The comparison of a new analytical system to one of known accuracy, using SLR and WLR, is reviewed. A previously reported technique for determining the merits of analytical systems, using only "raw" measurements, is reviewed and demonstrated far systems having constant SOs, RSOs. or bath.' The effect of transforming transmittance measurements to absorbances on the computations is examined. The software package is also used for descriptive statistics, Significance tests, and ANOVA. Many additional features, e.g., normality and outlier checks, residual analysis, and simulated data generation are demonstrated. The role of applied statistics and the microcomputer in chemometrics and the analytical laboratory is discussed.