Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2010

The application of two different dissimilarity matrices provides varying projections of species points in the plots of double principal coordinate analysis (DPCoA). In effect, this alters the locations of community points thus, affecting the ecological interpretations of datasets. With this gap, an opportunity was given to study the differences in the performance of DPCoA when different existing distance measures were applied in the construction of its dissimilarity matrix. Euclidean, maximum, fourth-order Minkowski, Canberra and Mahalanobis distance models and twenty datasets were the tools used in assessment. Five sets of Rao DIVCs and five species-based DPCoA projections were the generated for each data. Comparative analysis was done pair-wisely among the generated outputs through Spearman rank correlation coefficient and Procrustes analysis respectively. Results showed that the pairings between the Rao DIVC sets of Euclidean- maximum and Minkowski-based distance matrices obtained high correlation. Moreover, low sum of squared residual deviations on its corresponding DPCoA plots were outputted which further justified its resemblance. On the other hand, caution should be made usage of Canberra and Mahalanobis distance measures since both brought excessive changes on the values of Rao DIVCs and DPCoA results. Numerous measures were available for the analyst with some produce comparable results: however, the study cannot justify a preference since the process of assessment did not provide a basis for picking out the best model among the five measures tested. Nevertheless, the assessment made in the study showed and provided an insight that distance models like Euclidean, Maximum and Minkowski distance models were less sensitive in the application of DVCoA compared with that of the measures using Canberra and Mahalanobis.