Shaun S. Wulff

Robust Parameter Design Using Generalized Linear Mixed Models

In robust parameter design, it is often the case that the quality characteristic is nonnormal. An example in semiconductor manufacturing is resistivity, which typically follows a gamma distribution. It is also common to have models that contain, in addition to fixed polynomial effects, a random effect representing an extraneous source of variation. In this article, we demonstrate the use of generalized linear mixed models (GLMM) for situations in which the response is nonnormal and in which the noise variable is a random effect. We discuss the analysis of the random effects as well as the fixed effects in a fitted model using GLMM techniques. A numerical example from semiconductor manufacturing is provided for illustration.

An Overview of Optimization Formulations for Multiresponse Surface Problems

Many industrial applications involve more than one quality characteristic. For example, in robust parameter design, the quality characteristics include the process mean and process variance. Such applications lead to multiresponse surface problems in which it is necessary to determine optimal operating conditions according to some specified optimization criterion involving the quality characteristics. The purpose of this article is to address this problem from a multiobjective decision-making framework. The foremost approaches in multiresponse optimization are categorized and integrated. Guidelines are presented to help select appropriate formulations. Moreover, the applicability and computational aspects of the methods in various decision-making contexts are discussed. Numerical examples are also provided. Copyright

“Of Course I'm Communicating; I Lecture Every Day”: enhancing teaching and learning in introductory statistics

This article focuses on one instructors evolution from formal lecturing to interactive teaching and learning in a statistics course. Student perception data are used to demonstrate the instructors use of communication to align the content, students, and instructor throughout the course. Results indicate that the students learned, that communication in the alignment process played a role in the learning, and that the instructor used four key categories of communication: encouraging open communication, demonstrating examples interactively, structuring opportunities for application through problem-solving, and engaging students in reflection about their learning. The article concludes with the instructors reflections on how communication and the approach to teaching and learning changed as a result of conducting the scholarship of teaching and learning in the course.

Does Sex Make a Difference? Job Satisfaction of Television Network News Correspondents

Women have been employed in national television news since the 1960s, but there is a question about gender balance now. This study surveyed correspondents at national television news networks. The results showed that on average women were significantly younger and less experienced than men, which affected womens salaries. Women were less satisfied than men about their work environment, and after controlling for years employed at their network, women also were less satisfied with their jobs overall.

Estimation in Second-Order Models with Errors in the Factor Levels

The impact of errors in the factor levels is examined on the estimation of parameters in second-order response models. Errors can occur in setting the factor levels for response surface and robust parameter design models. These errors can lead to heterogeneity of variances in model errors that make ordinary least squares estimation inappropriate. Weighted least squares and maximum likelihood estimation approaches are developed as viable alternatives where it is assumed the variances and covariances of the errors are known. Performance of these estimation techniques are examined in simulation studies for two examples. Another example is given that applies these results.

Existence of maximum likelihood estimates in normal variance-components models

Abstract Under the usual nonnegativity constraints on the variance components, a maximum likelihood estimate (MLE) of the parameter vector in a normal variance-components model is known to exist. We investigate the question of existence under more general types of constraints and, in particular, under the constraint that requires only that the variance–covariance matrix be positive definite. Attention is restricted to models in which all the possible variance–covariance matrices commute with one another. It is found that in some models, such as all random one-way models with a single group having the largest size and all balanced random two-way models, the likelihood becomes infinite under the positive definiteness constraints, so that no MLE exists. In (practically) all normal balanced mixed-effects classification models, a residual maximum likelihood estimate (REMLE) exists.

Comparison of Parametric, Nonparametric and Semiparametric Modeling of Wind Tunnel Data

ABSTRACT This article presents a case study illustrating the use of parametric, nonparametric, and semiparametric modeling. The study comes from aeronautical research that examines a highly nonlinear relationship between coefficient of lift and angle of attack in a wind tunnel experiment. For this study, nonparametric and semiparametric approaches provide reasonable alternatives to the usual linear parametric approach and we demonstrate that they can aid in understanding the underlying process. Model comparison and replicated design points are also addressed in the article.

Resilient modulus of subgrade materials for mechanistic-empirical pavement design guide

This paper describes the test program, proposed test protocol for subgrade resilient modulus (Mr), determination of Mr values, and development of design tables and constitutive models for Mr estimations. Subgrade samples were collected from 12 locations throughout the state of Wyoming, USA for standard laboratory and Mr tests. The proposed Mr test protocol was validated using test data collected from a Round Robin Test Program. The study shows that Mr values of soils having R > 50 increase with increasing deviator stresses. In contrast, Mr values of soils having R ≤ 50 decrease with increasing deviator stresses. Design tables of Mr values were developed. Also, three constitutive models were locally calibrated for Mr estimations. The constitutive Model Type B yields a relatively better estimation of the resilient modulus than Models A and C. The proposed methods could be adopted to increase the efficiency of pavement design in the USA and other countries.

Time Series Analysis: Forecasting and Control, 5th edition

Time Series Analysis: Forecasting and Control, 5th edition by George E. P. Box, Gwilynm M. Jenkins, Gregory Reinsel, and Greta M. Ljung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Shaun S. Wulff Functional and Shape Data Analysis by A. Srivastava and E. P. Klassen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Li Xu and Yili Hong

Systematic back-calculation protocol and prediction of resilient modulus for MEPDG

Abstract A research focusing on the characterisation of representative local material properties was conducted to facilitate the full implementation of the Mechanistic-Empirical Pavement Design Guide for roadway designs in Wyoming. As part of the test program, falling weight deflectometer deflection data were collected from 25 test sites in Wyoming for back-calculation of subgrade resilient modulus. Also, subgrade materials from these test sites were sampled for laboratory resilient modulus measurement in accordance with the AASHTO T 307. The back-calculation is a user-dependent procedure and produces a non-unique resilient modulus estimation. To alleviate this limitation, this paper focuses on the recent development of a systematic back-calculation protocol for subgrade resilient modulus using MODCOMP6 software. The protocol is intended for use on a flexible pavement with a crushed base. The proposed procedure discusses pre-analysis checks, seed modulus adjustment, pavement structure adjustment and program termination criteria. A correlation study was conducted to correct back-calculated resilient modulus to laboratory-equivalent values. The results conclude that a non-zero intercept linear regression model provides a better correlation than the widely used zero intercept linear regression model. Furthermore, better correlations are achieved when the back-calculated resilient modulus of a lower subgrade layer and resilient modulus measured at higher laboratory test sequences Nos. 11 to 15 are considered. The non-zero model based on Mr test sequence No. 14 and lower subgrade layer yields the best correlation. For the zero model, a C-factor of 0.645 based on Mr test sequence No. 15 and lower subgrade layer yields the best correlation.