A1403
Title: Metric skewness for object data
Authors: Joni Virta - University of Turku (Finland)
Vida Zamanifarizhandi - University of Turku (Finland) [presenting]
Janne Kujala - University of Turku (Finland)
Oona Rainio - University of Turku (Finland)
Abstract: As datasets become increasingly complex, more advanced methods for extracting insight from data are required. Recently, analyzing data in metric spaces has gained attention and several descriptive statistics have been developed. However, important distributional characteristics, such as skewness, which provides valuable information about its structure, remain largely unexplored for object data in metric spaces. A novel method is introduced to compute the metric skewness. Moreover, its use in testing for the presence of skewness exhibits promising outcome both in level and power, surpassing its multivariate skewness counterparts in Euclidean space. Finally, the proposed metric skewness is applied as an inferential statistic to positron emission tomography (PET) data, illustrating its practical applicability.
