(eng): Statistical analysis and modelling of various phenomena are well established in nowadays industrial practice. However, the traditional approaches neglecting the true properties of the phenomena still dominate. Among others, this includes also the cases when a variable with bounded range is analyzed using probabilistic distributions with unbounded domain. Since many of those variables nearly fulfill the basic conditions imposed by the chosen distribution, the properties of used statistical models are violated rather rarely. Still, there are numerous cases, when inference with distributions with unbounded domain may lead to absurd conclusions. The paper addresses this issue from the Bayesian viewpoint. It briefly discusses suitable distributions and inferential methods overcoming the emerging computational issues.