Clustering medication compliance using rank profiles
LE3 .A278 2008
Bachelor of Science
Mathematics and Statistics
Mathematics & Statistics
Currently, there are few methods used in multivariate analysis to deal with missing values and many times researchers choose to discard observations that contain them. However, much information is lost when discarding observations that may contain only one or two missing values. Medication compliance datasets, due to their nature, are rife with missing values and are therefore rendered unusable in traditional clustering algorithms. There is a great desire to objectively group patients as compliant or non-compliant in order to study their characteristics and implement procedures in order to increase compliance. If medication compliance datasets tended to be complete, unsupervised classification procedures, such as heirarchical clustering, could be used in order to gain insight into drug-taking behaviours. By creating rank profiles across each patient and then utilizing rank correlation measures with missing values introduced by Alvo and Cabilio (1995), we will compute distance matrices with which clustering can continue on ‘as usual.’ In this way, heirarchical clustering will be applied to medication compliance datasets.
The author grants permission to the University Librarian at Acadia University to reproduce, loan or distribute copies of my thesis in microform, paper or electronic formats on a non-profit basis. The author retains the copyright of the thesis.