Applications of generalized additive models
Joshi, Miland (2011) Applications of generalized additive models. PhD thesis, University of Warwick.Full text not available from this repository.
Official URL: http://webcat.warwick.ac.uk/record=b2578185~S1
Main Purpose The study is primarily a contribution to a question of strategy rather than the development of a new method. It explores the circumstances in which the use of generalized additive models can be recommended. It is thus a contribution to answering the question "When is it a good idea (or not so good an idea) to use GAMs?" Content Following an introductory exposition in which they are compared to generalized linear models, subsequent chapters deal with evidence that could support possible recommendations: 1. A survey of recent studies, in which GAMs have been used and recommended, regarded with greater reserve, or compared to other methods. 2. Original case studies in which the applicability of GAMs is investigated, namely: (a) Receiver operating characteristic curves used in medical diagnostic testing, the associated diagnostic likelihood ratios, and the modelling of the risk score. (b) A study of a possible heat wave effect on mortality in London. (c) Shorter studies, including a study of factors influencing the length of stay in hospital in Queensland, Australia, and a simulation study. 3. Diagnostics, looking in particular at concurvity, and the problems of defining and detecting it. The study ends with recommendations for the use of GAMs, and possible areas for further research. The appendices include a glossary, technical appendices and R code for computations involved in the project.
|Item Type:||Thesis or Dissertation (PhD)|
|Subjects:||Q Science > QA Mathematics|
|Library of Congress Subject Headings (LCSH):||Linear models (Statistics)|
|Date:||5 December 2011|
|Institution:||University of Warwick|
|Theses Department:||Department of Statistics|
|Supervisor(s)/Advisor:||Shaw, J. Ewart H. ; Firth, David, 1978-|
|Extent:||387 leaves : ill., charts|
Actions (login required)