H-Type Indices With Applications in Chemometrics: h-Accuracy Index for Evaluating and Comparing Errors in Analytical Chemistry.

Abstract

Inspired by the popular h-index in bibliometrics, the h-accuracy index (HAI) was suggested as a new indicator to evaluate and compare errors in analytical chemistry. The HAI simultaneously considers the "trueness" of analytical measurements and the frequency of measurements with high "trueness". The HAI was defined as if the "trueness" of at most M% of the total measurements is no less than M%, the value of HAI will be M%, where a specific definition of "trueness" was given to compute the HAI. The range of the HAI was between zero and one. The HAI was applied to two examples as a new error index: (1) to evaluate and compare the analytical results by different methods/labs and (2) to evaluate and compare the prediction performances of different multivariate calibration models. Data analysis indicated that the HAI was a reasonable, robust, easy-to-compute, and comprehensive index for evaluating and comparing errors in analytical chemistry. Moreover, the HAI could provide the information about how many measurements are good.