Detailed Regression Models and Complete Set of Boxplots for Validation Step for the Paper: Multiple Imputation on Design of Experiments with Multiple Responses using STATA
The article contains the final regression models that complement the paper “Multiple Imputation on Design of Experiments with Multiple Responses using STATA” These final results are obtained after: (a) using the predictive mean matching (pmm) impute option under the command mi impute in STATA, and (b) adjusting significant linear regression models for each of the 15 response variables in the study using the command mi estimate. In this study, the mi impute pmm command was invoked to impute m=5 values for each response variable at each experimental condition with missing data. In this study, quadratic models are assumed to describe each response in terms of the 7 factors or independent variables. The initial regression models contained all the possible terms in a second order polynomial model on the factors. Quadratic terms and second order interactions were included except those involving categorical variables. The non-significant factors are iteratively removed until significant regression models are obtained for each response variable.
multiple imputation, design of experiments, multiple response optimization, statistics, Ingram School of Engineering
Novoa, C., Asiabanpour, B., & Alkusari, S. (2015). Detailed regression models and complete set of boxplots for validation step for the paper: Multiple imputation on design of experiments with multiple responses using STATA.