A nonparametric control chart for multivariate processes

In this paper, a nonparametric procedure for identification of errands in multivariate process control is presented. This approach combines two well studied nonparametric: techniques, namely the concept of data depth, and partitioning of the data space, The proposed procedure, calculates a depth measure, according to a reference in-control data set, for each cell on the partitioned space, Once the depth for each cell is calculated, a probability value is assigned to the cell. For any new point that comes from the analyzed process, classification for in or out-of-control is done based on the probability assigned to the cell containing the data, Results from simulated data are used to compare with the existing methodology.