+Computing adjusted projection depth using GSO algorithm

Projection depth is an important concept of non-parametric inference on multivariate data analysis and is the most widely used as statistical depth notion among all the existing depth procedures. This depth procedure was based on Stahel-Donoho multivariate location and scatter estimator and its outlyingness. Further, the adjusted outlyingness concept was used to compute projection depth and namely adjusted projection depth. In both the depth procedures, exact, fixed and random algorithms have been used to compute projection depth values. In this paper, the Gram-Schmidt Orthonormalization (GSO) algorithm is proposed to compute adjusted projection depth in order to improve the accuracy measure. The superiority of the GSO algorithm based adjusted projection depth over the exact, random and fixed algorithms has been demonstrated by applying it in the context of classification analysis by computing average misclassification error rate under real and simulation environments.