Book chapter
Identifying risk factors for cognitive change in the Women's Health Initiative: a neural networks approach
Hormones, Cognition and Dementia, pp.11-24
09/24/2009
DOI: 10.1017/CBO9780511635700.003
Abstract
Editors' introductionThe Women's Health Initiative Memory Study (WHIMS) has been much criticized and it has been suggested that subgroups of women exist for whom hormone therapy (HT) might improve or impair cognitive function. Possible modifying variables could be age, smoking, body mass index, and menopausal symptoms. These were included in artificial neural networks (ANN) analyses, which allow testing of complex non-linear higher order interactions of variables to predict outcomes. Artificial neural networks analyses without hidden units could predict responders and non-responders to treatment as well as logistic regression models that included only main effects, which indicated that higher order interactions were not necessary and did not add to the value of the models. There seemed to be no subgroups (e.g., older women who smoke and have a high body mass) for whom HT has a worse or better effect on cognitive function over time. This study also showed that cross-validation is essential in building robust models with many independent variables and should be applied as a standard technique in complex multivariate analyses.
Details
- Title: Subtitle
- Identifying risk factors for cognitive change in the Women's Health Initiative: a neural networks approach
- Creators
- Stephan BandelowMark A EspelandVictor W HendersonSusan M ResnickRobert B WallaceLaura H CokerEef Hogervorst
- Resource Type
- Book chapter
- Publication Details
- Hormones, Cognition and Dementia, pp.11-24
- DOI
- 10.1017/CBO9780511635700.003
- Language
- English
- Date published
- 09/24/2009
- Academic Unit
- Epidemiology; Injury Prevention Research Center; Internal Medicine
- Record Identifier
- 9984363578602771
Metrics
4 Record Views