The well-known statistical tests have been developed on the basis of many additional assumptions, among which the normality of a data source distribution is one of the most important. The outcome of a test is a p-value which may is interpreted as an estimation of a risk for a false negative decision i.e. it is an answer to the question “how much do I risk if I deny?”. This risk estimation is a base for a decision (after comparing with a significance level α): reject or not. This sharp trigger – p-level greater than α or not – ignores the fact that a context is rather smooth and evolves from “may be” through “rather not” to “certainly not”. An alternative option for such assessments is proposed by a fuzzy statistics, particularly by Buckley’s approach. The fuzzy approach introduces a better scale for expressing decision uncertainty. This paper compares three approaches: a classic one based on a normality assumption, Buckley’s theoretical one and a bootstrap-based one

The well-known statistical tests have been developed on the basis of many additional assumptions, among which the normality of a data source distribution is one of the most important. The outcome of a test is a p-value which may is interpreted as an estimation of a risk for a false negative decision i.e. it is an answer to the question “how much do I risk if I deny?”. This risk estimation is a base for a decision (after comparing with a significance level α): reject or not. This sharp trigger – p-level greater than α or not – ignores the fact that a context is rather smooth and evolves from “may be” through “rather not” to “certainly not”. An alternative option for such assessments is proposed by a fuzzy statistics, particularly by Buckley’s approach. The fuzzy approach introduces a better scale for expressing decision uncertainty. This paper compares three approaches: a classic one based on a normality assumption, Buckley’s theoretical one and a bootstrap-based one