Regression analysis, Rubber, Artificial
The goal of this study is to find a more organized and directed approach to build models for mixture systems. An attempt is made to generate and then compare Scheffe (mixture) models with those generated by McGee using the ‘conventional’ method for neoprene data. The models are judged on their ability to predict physical properties of neoprene by comparing the following: predicted and actual values by inspection; the calculated % error of prediction; the squared multiple correlation coefficients; adjusted squared multiple correlation coefficients; the Fisher statistic and significance probability. Scheffe models do not have an intercept term and test statistics which appear on the computer printout are inflated. Pseudocomponents and Scheffe-equivalent models are procedures used to obtain accurate test statistics to describe the selected Scheffe models. The effectiveness of these two procedures is evaluated. Results indicate that Scheffe models are better predictors for the physical properties of neoprene than those generated by McGee using the ‘conventional’ method in 1980. Scheffe-equivalent equations are found to be more reliable than pseudocomponents for generating accurate test statistics to describe the selected Scheffe models.
McGee, William W.
Master of Science (M.S.)
College of Arts and Sciences
Length of Campus-only Access
Masters Thesis (Open Access)
Loh, Cheng Y., "The Use of Scheffe-Equivalent Equations to Predict Physical Properties of Neoprene" (1986). Retrospective Theses and Dissertations. 4895.