Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.569125
Title: Multivariate modelling of cognitive function and brain structural data
Author: Cheyne, Christopher Paul
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2011
Availability of Full Text:
Access through EThOS:
Abstract:
Previous studies have investigated links between cognitive ability and a number of factors including age, gender, handedness, musical ability as well as the volume and surface area of certain brain structures, However, in these studies either the explanatory variables are analysed independently of each other, or the investigation is based on a separate analysis for individual cognitive outcomes (e.g. language, visuospatial, etc.) The main objectives of this thesis are(1) to develop general multivariate models, which include mixed-effects terms, to account for the correlation in the data, (2) to explore the possible associations in children and adults between multiple cognitive ability test scores and the range of factors mentioned above by simultaneously applying the multivariate models designed in (1), and (3) to investigate the possible effects of missing data on the results. To meet these objectives, a range of statistical and stereological methods was employed: Multivariate linear and linear mixed models were developed and fitted to multiple datasets. The approach used took into account the correlation of clustered data, the correlation between outcomes as well as the association between explanatory variables and a linear combination of the outcomes. Stereological methods were used to estimate the volume and surface area of a region of the brain called Broca's area, using magnetic resonance images. Also, the latest formulae in error prediction for these stereological estimates were described and applied to the data. Results from the fitted multivariate linear mixed model to a dataset of l l-year old children (n= 1184 3) showed that children whose writing hand has less hand skill than the opposite hand performed worse, on average, in both reading and maths scores, than those children whose writing hand had more hand skill than the opposite hand. A multivariate linear model fitted to a dataset of adults (n=142) revealed that the gender difference found in the non-musician groups for the vocabulary and arithmetic scores was not present in the musician group. Multivariate linear models were subsequently fitted to a subset of this cohort containing volume and surface area estimates of Broca's area (n=39). Musicians were associated with Broca's area being less convoluted in the right hemisphere than non-musicians. Other associations investigated were not found to be statistically significant. Inverse probability weighting was then used to take the missing data into account for each of the analyses (aim (3)). The results and interpretations determined from the fitted multivariate models were consistent with the analyses when the missing data were accounted for. Only those results for the children dataset changed slightly, but not enough to alter the interpretations of the results. This adds weight to the belief that the results of the multivariate analyses gave a reasonably accurate description of the variability that exists within the children and adult datasets.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.569125  DOI: Not available
Share: