| dc.contributor.author | Hughes, Stephen Gerard. | en_US |
| dc.date.accessioned | 2014-10-21T12:37:22Z | |
| dc.date.available | 1996 | |
| dc.date.issued | 1996 | en_US |
| dc.identifier.other | AAINN15833 | en_US |
| dc.identifier.uri | http://hdl.handle.net/10222/55109 | |
| dc.description | Solutions to modern problems in analytical chemistry often require the use of multivariate data sets in which the measurements for individual samples consist of a vector or matrix. While present day instruments make the acquisition of such data relativity simple, the challenge has become presenting it in an informative and logical manner. Chemometrics is the branch of analytical chemistry that addresses this challenge. | en_US |
| dc.description | When treating multivariate data sets, the method of data analysis depends greatly on the underlying structure of the data set. It is found, for example, that the presence of "order" in data sets can greatly simplify analysis. An ordered data set is one in which the component contributions change systematically in accordance with some underlying ordinal variable (e.g. time, pH). In this work, a method for the analysis of such ordered data sets, called evolving projection analysis (EPA), is extended to mixtures of more than two-components and to systems with nonideal detector response. Applications to liquid chromatography and spectrophotometric titrations are considered. | en_US |
| dc.description | In contrast, the sequence of samples in disordered data sets is essentially random. A number of standard methods have been devised for these kinds of data sets and this work considers an interesting case involving conductivity prediction of non-brine water samples using a variety of calibration methods. Improvements to these methods and the limitations of this approach are discussed. | en_US |
| dc.description | Finally, methods were sought to search for order in disordered data sets, working from the hypothesis that many data sets that appear disordered are actually ordered, if only the proper ordinal variable could be identified. It was found that the ordering algorithm, implemented using the Genetic Algorithms could be successfully applied to problems involving cluster analysis and evolving data sets. A novel study of environmental receptor modelling data showed that this approach can reveal unique information hidden in the disordered data set. | en_US |
| dc.description | Thesis (Ph.D.)--Dalhousie University (Canada), 1996. | en_US |
| dc.language | eng | en_US |
| dc.publisher | Dalhousie University | en_US |
| dc.publisher | | en_US |
| dc.subject | Chemistry, Analytical. | en_US |
| dc.title | Chemometrics for ordered and disordered data sets. | en_US |
| dc.type | text | en_US |
| dc.contributor.degree | Ph.D. | en_US |
