Discrete directional data exhibiting p-fold rotational symmetry, where directions separated by integer multiples of 2π/p are equivalent, arise in various applications but are often modeled using continuous distributions. In this article, we develop two general methods for constructing distributions for such data: symmetrization and folding of discrete circular distributions, and wrapping of discrete linear distributions. Particular models arising from these constructions are studied, together with their properties and estimation methods. For axial sand-grain orientation data, the proposed models are compared with discretized continuous axial distributions, whose category probabilities require numerical integration. The proposed models attain smaller Akaike information criterion values among the models considered while allowing direct likelihood evaluation without numerical integration. The framework is also applied to binned Manhattan street-orientation data, demonstrating its applicability to higher-order rotational structures, specifically the 4-fold case.

翻译：暂无翻译