In color-imaging science researchers frequently conduct experiments that are evaluated by measuring a number of ΔE values. Often the effect of various experimental parameters is evaluated by comparing the means of two or more sets of ΔE values. Frequently, a t-test is used but correct application of the t-test requires that a number of assumptions (such as the data being normally distributed or the sample size is big enough) are satisfied. In this article, an improved T statistic has been used in a case study where the data are skewed and do not have a normal distribution while at the same time the sample size is small. The importance of this assumption comes to mind when we realize that most of the derived data from the color-imaging science are not applicable for t-test because of not following a normal distribution.
Mohsen Mohammadzadeh, Maryam Mohammadzadeh, Stephen Westland, "The Significance Testing of a Skewed Color-Imaging Data Set" in Proc. IS&T 19th Color and Imaging Conf., 2011, pp 43 - 47, https://doi.org/10.2352/CIC.2011.19.1.art00010