dc.contributor.advisor	Li, Longhai
dc.contributor.advisor	Xing, Li
dc.contributor.committeeMember	Li, Junxin
dc.creator	Hu, Hao
dc.date.accessioned	2022-09-29T19:45:39Z
dc.date.available	2022-09-29T19:45:39Z
dc.date.copyright	2022
dc.date.created	2022-09
dc.date.issued	2022-09-26
dc.date.submitted	September 2022
dc.date.updated	2022-09-29T19:45:39Z
dc.description.abstract	Background: Numerous studies have shown that older adults’ cognitive abilities are age-related and likely to decline at a certain age. Based on these indications, this work uses functional principal component analysis (FPCA) to explore changes in cognitive function with age. This study aims to describe the longitu- dinal cognitive function of elderly trajectory patterns between 65 and 80 years of baseline age using FPCA and identify risk factors for cognitive decline using machine learning algorithms. Methods: We used FPCA to extract the overall pattern change and use elastic-net, decision tree, and random forest models to find risk factors. In a sample of elderly (n = 944) with 6608 measurements (7 waves) from the Survey of Health, Aging, and Retirement in Europe (SHARE), by using age at interview as time, longitudinal cognitive function trajectory patterns for the elderly were extracted using FPCA. Zou and Hastie (2005) proposed the elastic net regression method, which effectively implements feature selection by setting coefficients of non-significant variables to zero [56]. Random forest is a tree-based machine learning algorithm that harnesses the power of multiple decision trees to make decisions. Random forests combine the outputs of individual decision trees to generate the final result. [15]. We modelled the first two functional principal components (FPCs) with these machine learning algorithms and used the selected covariates in the baseline wave to identify risk factors. Results and Conclusions: We have obtained four FPCs explained by 78.0, 14.2, 6.7 and 1.1 % of the variation respectively for the cognitive function. The mean function of FPCA shows that cognitive decline is generally divided into two stages (early decline and late decline). By analyzing and comparing a set of models at the national level, the cognitive function of each country is slightly different. Older people in Italy and Spain have significantly lower cognitive abilities. The Predictive R2 of FPC1 and FPC2 is around 0.5 and 0.2 with covariate delayed recall score (DRS) and reduced to 0.4 and 0.1 without covariate DRS. From the individual point of view, many risk factors are modifiable and can be prevented in advance. Our results show that immediate recall score, education level, country, numeracy score, reading score, and household income are associated with cognitive patterns in the elderly.
dc.format.mimetype	application/pdf
dc.identifier.uri	https://hdl.handle.net/10388/14244
dc.language.iso	en
dc.subject	FPCA
dc.subject	cognitive decline
dc.title	Identifying Risk Factors for Cognitive Decline Using Statistical Learning Techniques and Functional Data Analysis
dc.type	Thesis
dc.type.material	text
thesis.degree.department	Mathematics and Statistics
thesis.degree.discipline	Mathematics
thesis.degree.grantor	University of Saskatchewan
thesis.degree.level	Masters
thesis.degree.name	Master of Science (M.Sc.)

Identifying Risk Factors for Cognitive Decline Using Statistical Learning Techniques and Functional Data Analysis

Files

Original bundle

License bundle

Collections