Estimating values of a quality characteristic after product conversion

Abstract

“Estimating Values of a Quality Characteristic After Product Conversion” is a statistical study commissioned by a company in the United States. Due to the proprietary nature of the information contained within this study, all proper names have been changed to a generic form and all references to actual company personnel and documentation have been removed. At a production facility in Anytown, U.S.A., Company A makes product B on production line C. Product B comes in a variety of colors, custom-converted widths and types. Throughout the production process many of the product’s physical attributes are monitored. Characteristic D is one such physical attribute. Characteristic D is measured, in E units, before product conversion and after product conversion. Personnel that work closely with product B had hypothesize that a correlation between pre-converted and post-converted test values exists. The problem of the study is to determine if there is a relationship between characteristic D values of product B created on production line C before and after conversion. If a quantifiable relationship exists, then post-converted test values could be predicted and post-converted testing eliminated, creating greater economies of operator labor and test equipment use. To solve the problem of the study, background information was gathered on the product and its characteristics and the test equipment and its operation. Data collection was planned for by learning about and utilizing Microsoft™ Excel and Microsoft™ Access and by remedying inaccurate data in the database. Data was collected and organized by creating and running a series of queries. Data was analyzed through the use of scatter plots, histograms, x-bar charts, descriptive statistics and corresponding rules and theorems. Regression and correlation analysis indicated a loose correlation between pre-converted and post-converted test values, but the correlation was not significant enough to merit further modeling. Instead, the Empirical rule and sample statistics of the average value and standard deviation of the change in characteristic D were used to predict future changes in characteristic D. Based on the analysis of the data, recommendations were made to discontinue post-converted characteristic D testing if pre-converted test values met or exceeded calculated lower specification limits.