A Bayesian Plotting Method for Fractional Factorial Data Analysis

Abstract

This paper proposes a plot-based method for fractional factorial data analysis. The proposed plot is called a “response-probability model analysis plot” (RPMAP) because it displays the predicted responses associated with alternative models and decisions versus the model posterior probabilities. Benefits of the proposed method include unique information about whether the current state of model uncertainty and achievement of objectives warrants additional experimentation. Also, in some cases, the RPMAP leads to settings with arguably superior robustness to model uncertainty compared with normal probability plots or Posterior probability plots. The methods are illustrated using the well-studied injection molding data and another real-world case study. In both cases, new insights are gained with potential value to practitioners.

• Design of Experiments
• Mixture Model
• Model Analysis Plot
• Optimization
• Screening Experiment

Additional information

Notes on contributors

Theodore T. Allen

Dr. Theodore T. Allen is an Associate Professor of Integrated Systems Engineering at The Ohio State University. He is a fellow of ASQ. His email address is [email protected].

Ravishankar Rajagopalan

Dr. Ravishankar Rajagopalan is a Lead Engineer at GE Energy, Bangalore, India. His email address is [email protected].