Statistical Process Control In Complex Data

Since Walter Shewhart introduce the Shewhart control chart in 1925,control charts increasingly have important applications in quality detection,health care and finance fields.The relative classic control charts include Shewhart control chart,Exponentially Weight-ed Moving Average(EWMA)control chart and Cumulative Sum(CUSUM)control chart.EW-MA control chart has a good detection performance for small and medium shifts because of considering the history sample and the current sample information.In order to solve some problems of EWMA control chart in complex data and further improve the EWMA control chart detection efficiency in the fields of health care and manufacturing processes,based on EWMA control chart,this paper do the following two aspects studies:(1)For the risk adjusted unit surgical data,this paper discussed a combined risk adjusted EWMA(RAEWMA)and VLAD control charts,namely RAEV control chart.The RAEV control chart has the statistical properties of RAEWMA and VLAD control charts.It not only provides a qualitative mechanism for identify whether a significant shift occurs but provides a quantitative mechanism for the net number of lives saved(or lost)at any moment.Finally,this paper combines a hospital surgery data in the Surgical Outcome Monitoring and Improvement Program(SOMIP)to illustrate the good statistical properties of the RAEV control chart.(2)For the unknown distribution and multivariate data,the observations of traditional multivariate EWMA control charts are assumed from a known normal distribution.However,in many practical monitoring processes,datasets are multivariate and the distribution are usu-ally unknown.Some researchers proposed multivariate control chart by using data depth.Data depth can deal with multivariate data problems effectively.Based on the proposed Mahalanobis depth,this paper further studies the adaptive multivariate nonparametric EWMA(MANE)con-trol chart.Finally,based on the provided MANE control chart,this paper combines the method of variable sampling interval(VSI)to further improve the detection efficiency of MANE con-trol chart.This control chart is called VSI-MANE control chart.Thus the VSI-MANE control chart can adaptive choose the sampling interval according to the statistics whether falling into a "warning zone" and detects shifts in a process as soon as possible.In addition,this paper compared the detection efficiency of proposed VSI-MANE control chart with some existing control charts.The results find the VSI-MANE control chart can detect the occurrence of shifts faster.Meanwhile,the paper also combines two real examples to illustrate the VSI-MANE control chart has a good detection efficiency.