Monitoring of Conway-Maxwell-Poisson profiles under multicollinearity

As a result of the application of evolving technologies in the industry, the data generated in the manufacturing process necessitates the development of new methods for process control which can address several problems simultaneously. In this thesis, we aim to identify shifts in the count profiles while reducing the negative impact of the multicollinearity caused by correlated predictors. We propose PCR, ridge, Liu, and r-k deviance-based Shewhart, CUSUM, and EWMA control charts for detecting out-of-control observations where the process data is modelled through the Conway-Maxwell-Poisson profile. The Conway-Maxwell-Poisson distribution facilitates the analysis of the count process data that exhibit varying levels of dispersion, whilst adjusted PCR, ridge, Liu, and r-k class estimation methods overcome the multicollinearity. Extensive simulation studies and real-life data analysis are carried out to illustrate the effectiveness of the proposed residual-based control charts.