Application of Semivariance in Uncertainty Interval Estimation for Road Noise Indicators - A Novel Method

The paper presents the author's method for constructing asymmetric uncertainty intervals based on a generalization of Chebyshev's theorem using semivariance. The paper compares the proposed confidence intervals using semivariance with classical confidence intervals for long-term noise indicators, Ld, Le, Ln, Lden. The data for comparison were values of equivalent daily sound levels from continuous road noise monitoring of the city of Gdansk from the period of a calendar year. The study showed that intervals based on semivariance provide a higher level of coverage of the actual value of noise indicators compared to classical confidence intervals, especially for small measurement samples and in cases where probability distributions are characterized by deviations from normality. Analyzed for each sound level indicator for data on sample sizes ranging from 5 to 50 elements showed that classical confidence intervals often fail to achieve the expected 95% coverage level, especially for distributions with high skewness. In contrast, intervals based on semivariance more effectively account for distribution asymmetry, offering a more reliable assessment of measurement uncertainty. The results suggest that the use of semivariance-based intervals may offer a better alternative to classical methods when dealing with non-standard distributions of measurement data.