Reading resolution is a well-defined, readily known
characteristic of the measuring instrument concerned. As real
variability is also affected by other factors, direct use of
reading resolution as bin width would lead to an over-detailed
description. A practical approach connecting such a readily
available information as reading resolution, with a reasonable
description of the distribution, is provided by the method of
kernel density estimation [9], a non-parametric method closely
related to histograms offering however additional advantages,
such as smoothness and continuity. Given a random sample
X1,…, Xn with a continuous, univariate density f, the kernel
density estimator is