摘要:

This article presents an extension of the response surface approach to Taguchi's parameter design by accommodating generalized linear modeling. The approach of Myers, Khuri, and Vining is taken as a starting point, treating noise variables as fixed effects. It is, however, assumed that process variability is caused not only by the presence of noise factors but also by residual variation. Therefore, in addition to a model with a linear link function that relates the process mean to design and noise factors, a second model with a log-link function is proposed that relates the variance to a linear predictor in the design factors. The conditional response model, with noise factors taken as fixed, is fitted by iteratively reweighted least squares. Unconditionally, response surfaces for the process mean and process variance are produced through the use of a mean and variance operator on the conditional response model. As an illustration, data from an injection molding experiment are reanalyzed and residual dispersion effects, conditional on the noise factors, are modeled.

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484548

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

I extend the concept of partial least squares (PLS) into the framework of generalized linear models. A spectroscopy example in a logistic regression framework illustrates the developments. These models form a sequence of rank 1 approximations useful for predicting the response variable when the explanatory information is severely ill-conditioned. Iteratively reweighted PLS algorithms are presented with various theoretical properties. Connections to principal-component and maximum likelihood estimation are made, as well as suggestions for rules to choose the proper rank of the final model.

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484549

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

Modeling of system lifetimes becomes more complicated when external events can cause the simultaneous failure of two or more system components. Models that ignore these common cause failures lead to methods of analysis that overestimate system reliability. Typical data consist of observed frequencies in whichiout ofm(identical) components in a system failed simultaneously,i= 1,…,m. Because this attribute data is inherently dependent on the number of components in the system, procedures for interpretation of data from different groups with more or fewer components than the system under study are not straightforward. This is a recurrent problem in reliability applications in which component configurations change from one system to the next. For instance, in the analysis of a large power-supply system that has three stand-by diesel generators in case of power loss, statistical tests and estimates of system reliability cannot be derived easily from data pertaining to different plants for which only one or two diesel generators were used to reinforce the main power source. This article presents, discusses, and analyzes methods to use generic attribute reliability data efficiently for systems of varying size.

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484550

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

摘要:

A new method of error-rate estimation in linear discriminant analysis appropriate to situations with several small groups is introduced. It is compared with three other methods that can be generalized from the two-group situation. Simulation studies show that in most situations it outperforms the other techniques.

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484551

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484552

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484554

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484556

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484557

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484560

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor

ISSN:0040-1706    

DOI:10.1080/00401706.1996.10484561

出版商:Taylor & Francis Group    

年代:1996

数据来源: Taylor