Akademik Veri Yönetim Sistemi
IV. International Applied Statistics Conference (UYIK - 2023) Sarajevo / Bosnia and Herzegovina, 26-29 September 2023, Sarajevo, Bosna-Hersek, 26 - 29 Ekim 2023, ss.165-166
A key issue in statistics is normality testing. Given that many statistical techniques assume normality,
it is crucial to ascertain if a dataset is normally distributed. But in high-dimensional environments,
conventional normality tests are sometimes unreliable, especially when the data has a generalized
Gaussian distribution. This study suggests a brand-new normality test for high-dimensional data that
is resistant to outliers, such as extended Gaussian distributions. The suggested test is based on a
nonparametric method called the nearest neighbour approach, which makes no assumptions about the
distribution of the data. The suggested test is applicable for huge datasets because it is computationally
effective as well. Both simulated and actual data are used to evaluate the proposed test. The findings
demonstrate that, even when the data has a generalized Gaussian distribution, the suggested test is
more effective than conventional normality tests in high dimensions. This piece significantly advances
the subject of statistics. For researchers that need to verify for normality in high-dimensional data, the
proposed test is a useful tool. The test is computationally effective and resilient to deviations from
normality, including modified Gaussian distributions. This makes a variety of applications possible.