Evaluation of the Performance of a Random Coefficient Regression Model CUSUM Control Chart Under Varying Model Conditions, with Human Services Applications

• Activity Number: 235
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Quality and Productivity
• Author(s): Christopher John McKinney*+ and Jay Schaffer
• Companies: University of Northern Colorado and University of Northern Colorado
• Keywords: CUSUM ; random coefficient regression model ; regression control chart ; human services

Abstract

The use of quality control charts with metrics within the educational, behavioral, and other human services field has typically been considered very limited due to the complications imposed by nested structures, conditional relationships, and complex variance-covariance structures. The current study demonstrates how under these population conditions the random coefficient regression model control chart (RCRMCC) is a useful alternative traditional control charts. Derived from the Regression Control Chart (RCC), the RCRMCC utilizes the residuals of the random coefficient regression model to provide the inputs for the quality control charts. Under the CUSUM framework and using Monte Carlo simulations, the RCRMCC and RCC are compared in regards to their ARLS under varying in-control and out-of-control population conditions. It was found that the RCRMCC is advantageous under conditions similar to those found in the educational, behavioral, and other human services industries. Applications of the findings and future directions with the RCRMCC are further discussed.