No label defined (Q4231189)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 1260618
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | No label defined |
scientific article; zbMATH DE number 1260618 |
Statements
11 March 1999
0 references
0.8529981970787048
0 references
0.8280449509620667
0 references
This book presents a thorough treatment of the concepts and methods of linear model analysis and illustrates them with numerical and conceptual examples. A substantial portion of statistical analysis is done in the context of a linear model that describes the data. There is an extensive literature on the methods of analyzing data that is associated with such models. Yet frequently we encounter a lack of agreement on the appropriate analysis, confusion over the application of the existing methods, and improper interpretation of the conclusions. The ready availability of high-speed computers and statistical software encourages the analysis of ever larger and more complex problems while at the same time increases the likelihood of improper usage. At least part of the problem lies in the gap between the developers of the methodology and the users of those methods. Computer software may fill the gap, but it may not accurately portray the method or it may be improperly interpreted by the user. One of the aims of this book is to bridge that gap by presenting the methods in a conceptually simple way so that the user will more easily understand the application of the methods and be able to assess whether the computer software reflects the method. This book is divided into three parts. Part 1 contains an introduction to the formulation of linear models, a discussion of the statistical distribution theory for linear and quadratic forms, a presentation of the basic linear model theory, and a review of the concepts and methods of simultaneous inference. Part II addresses linear regression models. The discussion focuses on the development of a model that properly reflects the data and the experimental situation. The problems associated with collinear predictors are discussed and illustrated and remedial solutions are examined. The literature on influential data diagnostics is reviewed and examined. Graphical methods for examining the data and the model are stressed and illustrated. A careful treatment is given to the topic of indicator variables. The ideas are developed in such a way as to avoid the frequent misuses of this useful concept. Part III is concerned with what are commonly called analysis of variance models. Three chapters are devoted to the discussion of fixed effects models. This methodology is concerned with the analysis of means of populations, and the cell means model is used as the basis for interpretation and hypothesis formulation. The cell means model, introduced by the author and \textit{F. M. Speed} [see J. Am. Stat. Assoc. 70, No. 351, 706-712 (1975; Zbl 0314.62030)], has now been commonly accepted in the literature. The discussions further illustrate the utility of this model formulation and describe the relation to the classical effects and interaction models. The model is especially powerful in the analysis of unbalanced data, including the problem of missing cells. The last three chapters of Part III are devoted to the mixed model problem, that is, models with a nonscalar covariance structure. Classical methods of estimation and inference are reviewed and a new approach for estimating variance components, called AVE estimation, is introduced. This method develops the estimates in a way that allows for a diagnostic analysis of the data as well as an assessment of the model and the design. The source of negative estimates of variance components is revealed. The method is especially useful if the objective of the experiment is to identify sources of variability in the process being studied. The method is initially applied to balanced data situations and is then extended to include the unbalanced case. The EM algorithm is used for the extension. In addition to the estimation and diagnostic analysis, the method suggests a general approach for making inferences on both the fixed and the random effects in the model.
0 references