Theoretical verification that multi-point measurement of infrared spectra will lead to high accuracy of discrimination between heterogeneous samples.

Abstract

If samples are compared with each other without averaging, and individual measurement values in a micro-region are left, there is a possibility that samples can be identified with higher accuracy. Therefore, in this study, we verified it using numerical calculation software (Matlab). I generated two groups of 100,000 random numbers that have different averages. Then, for both data groups, the number of data was reduced to various quantities using an averaging method. Next, both groups were compared by t-test and f-test. The probabilities of being different populations in both tests were almost constant even when the number of data changed. Therefore, it has become clear that increasing the amount of data does not improve accuracy.