Joint Research Conference

June 24-26, 2014

NOTE: The short course will take place in the Wallingford Room at the Watertown Hotel.  It will begin at 8:30 and a continental breakfast will be served prior.  Lunch is also included.

On Monday, June 23, 2014, the Quality & Productivity section of ASA will offer a short course:

An Introduction to Bayesian Statistical Process Control

By: Panagiotis Tsiamyrtzis, Athens University of Economics & Business

In the field of statistics the dilemma of being frequentist or Bayesian reflects the underline philosophy of the person performing the analysis. While, in almost all areas of statistics, frequentist and Bayesian methods coexist, in Statistical Process Control (SPC), the frequentist’s approach seems to dominate the field.

In general, despite the fact that the frequentist SPC methods work efficiently, there are several cases where they encounter difficulties and they need to impose restrictive assumptions. Also, the frequentist based inference is quite often misinterpreted from non-specialists. Finally, prior information (usually available from earlier experience with the same or a similar process and/or expert’s opinion) regarding the process is left unexploited.

In this short course, we will introduce and motivate the use of Bayesian methods in standard SPC practice. The flexibility of the Bayesian approach will allow us to relax some of the frequentist’s assumptions and the decision theory will help to provide easy to interpret inference. Most importantly though, the Bayesian approach will provide an axiomatic framework where prior information can be utilized allowing immediate inference.

The short course will consist of two parts. In the first one the fundamentals of the Bayesian approach, along with its computational aspects will be presented. In the second part, we will review various Bayesian methods that can be employed in the Bayesian SPC framework and we will provide Bayesian alternatives of existing frequentist SPC tools. The Bayesian methods will be motivated by examples, where the frequentist’s approach encounters difficulties (like phase I management and short runs).

No prior knowledge regarding the Bayesian approach is required and only basic familiarity with statistical theory and standard SPC methods will be helpful.

Course Outline:

Introduction to Bayesian statistics
Likelihood/Frequentist/Bayesian approach
Bayes theorem
Prior distribution
Sequential updating
Sensitivity Analysis
Decision Theory (Bayes/Minimax rules)
Inference (point/interval estimation and hypothesis testing)
Predictive Inference
Multivariate Bayesian statistics
Markov Chain Monte Carlo (MCMC) methods
Bayesian SPC methods
Kalman Filter
Bayesian change point methods for parameter monitoring
Quality Measurement Plan

Pre-Conference Short Course

June 23, 2014