Using a two-sided power distribution family in processing measurement results

When processing measurement results and evaluating their accuracy, a normal distribution is traditionally used; however, in some cases this approach may be incorrect (for example, if the distribution is significantly asymmetric or it is bounded). Therefore, the selection of a family of distributions is relevant, which would allow the approximation of both traditionally applied normal and uniform laws, as well as encompass distributions of other types that more adequately reflect the specifics of particular measurement tasks. To model measurement data and evaluate accuracy indicators, it is proposed to use the family of two-sided power distributions. The choice of this family is due to the fact that it is described by an extremely simple mathematical model while providing sufficient variety of probability density function shapes, including uniform, bimodal, and unimodal distributions (both symmetric and asymmetric). Methods for fitting (assigning) the distribution law are provided based on both a priori information (using specified accuracy indicators: measured value, standard or expanded uncertainties, coverage interval boundaries for a given probability level) and sample data (including maximum likelihood, the method of moments, and two-sided power distribution fitting via inverse transformation). Metrologically justified criteria for the approximation of continuous distributions by distributions of the family under consideration are proposed, and its application in evaluating the uncertainty of multiple equally precise measurements is described. Software has been developed for transforming distributions, which allows the measurement model to be specified in an analytical form. The results obtained are useful for specialists applying statistical modeling methods when evaluating measurement results, certifying measurement procedures, and processing data from interlaboratory comparisons.

1. Possolo A., Merkatas C., Bodnar O. Asymmetrical uncertainties. Metrologia, 56(4), 045009 (2019). http://dx.doi.org/10.1088/1681-7575/ab2a8d

2. Sahlin E., Magnusson B., Svensson T. Calculation of uncertainty intervals for skewed distributions – Application in chemical analysis with large uncertainties. RISE Report 2021:07. RISE Research Institutes of Sweden, Gothenburg, Sweden (2021). http://dx.doi.org/10.13140/RG.2.2.27781.22241

3. Cowen S., Ellison S. Reporting measurement uncertainty and coverage intervals near natural limits. The Analyst, 131(6), 710–717 (2006). http://dx.doi.org/10.1039/B518084H

4. Kotz S., Van Dorp J. R. Beyond Beta: Other Continuous Families of Distributions with Bounded Support and Applications. World Scientific Publishing, Singapore (2004). https://doi.org/10.1142/5720

5. Kotz S., Van Dorp J. R. The Standard Two-Sided Power Distribution and its Properties, The American Statistician, 56(2), 90–99 (2002). https://doi.org/10.1198/000313002317572745

6. Zolotarev V. M. Modern theory of summation of independent random variables. Nauka, Moscow (1986). (In Russ.)

7. Stepanov A. V., Chunovkina A. G. On the estimation of parameters of the asymmetric TSP distribution and its application. Probabilistic methods in discrete mathematics: abstracts of reports of the XI international conference, Petrozavodsk, Karelian Research Center of the RAS, рр. 106–108 (2024).

8. Stepanov A. V. On estimating the coverage interval of a standard two-sided power distribution from sample data. Contemporary Mathematics. Fundamental Directions, 71(2), 341–352 (2025). (In Russ.) https://doi.org/10.22363/2413-3639-2025-71-2-341-352

9. Tyrsin A. N. Method for selecting the best distribution law for a continuous random variable based on inverse mapping. South Ural State University Bulletin. Series: Mathematics, Mechanics, Physics, 9(1), 31–18 (2017). (In Russ.). https://doi.org/10.14529/mmph170104

10. Stepanov A., Chunovkina A. On the Estimation of TSP Distribution Parameters Based on Experimental Data. 2025 XXVIII International Conference on Soft Computing and Measurements (SCM), Saint Petersburg, Russian Federation, 28–30 May (2025). https://doi.org/10.1109/SCM66446.2025.11060308