Bibliography: leaves 223-229. / xiii, 229 leaves : ill. ; 30 cm. / Title page, contents and abstract only. The complete thesis in print form is available from the University Library. / This thesis considers several aspects of multivariate analysis motivated by the South Australian prenatal Down's syndrome screen, which seeks to identify those pregnancies at high risk to Down's syndrome from several risk factors, amongst them maternal age and measurements from four analyte concentrations found in maternal blood. The primary aim is to discriminate between the unaffected and affected pregnancies in terms of these observed risk factors. Through the development of multivariate regression models it is examined how the Mahalanobis distance might be decomposed into distances that give more insight into where the discrimination is derived and how this might change with the gestational age at which the analyte concentrations are measured. A second multivariate aspect motivated by the screen surrounds the detection of directional outliers and the incorporation of knowledge of likely departures into the outlier identification technique. A method is also developed to account for misreporting in inferences of gestational age. Similarities with measurement error and misclassification models are explored. / Thesis (Ph.D.)--University of Adelaide, Dept. of Statistics, 1998?
| Identifer | oai:union.ndltd.org:ADTP/260468 |
| Date | January 1998 |
| Creators | Henderson, Brent |
| Source Sets | Australiasian Digital Theses Program |
| Language | en_US |
| Detected Language | English |
Page generated in 0.0019 seconds