The Influence Study of Statistical Uncertainty on Probabilistic Outlier Detection for Geotechnical Engineering Data

Outlier is attached importance in statistics and engineering, because it might result misleading identification results. However, there significant uncertainty the outlier detection, when an outlying observation lies close to boundary between outliers regular data or are sparse observations. The associated of mostly results from statistical parameters, such as mean value standard deviation. unknown how influences detection. This paper compares two detection methods for influence study on probabilistic One based Mahalanobis distance (MD) using total probability theorem combining with half-means method (RHM). other RHM Bayesian machine learning (BML), which can consider uncertainties parameters MD. simulated dataset used comparative study. Different dimensional various numbers observations simulated. Thereinto, through double-mode triangle distribution. show that necessary multivariate