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

Limited studies have been conducted on the importance of choosing the right way of transforming a set of data in a manner that it can express its characteristics more or less strongly. An application of a specific data transformation to an original data matrix before carrying out PCA which provides varying projections of species points on its biplots. In effect, it alters the ecological and mathematical interpretation. Thus, an opportunity was recognized to study the performance of PCA-PCA Co-Inertia Analysis whenever varying data transformations were applied to the original data matrix before carrying out non-centered PCA. Logarithmic, centering, normalizing and square root transformations were utilized in the assessment. With each of the transformed data sets paired with the unaltered data matrix, twenty five PCA projection and fifteen Co-Inertia projections were generated. Pair-wise comparative analysis was done among the generated outputs through Procrustes Analysis. Results showed that low Procrustes statistic values were generated by the centered and square-root transformed data sets. Evidently, results also showed that Co-inertia projections using centering by site and normalization by site greatly considered dominant species in the data. Meanwhile, lesser weights were given to abundant species when the logarithmic and square root transformations were utilized. The list of data transformations utilized was not exhaustive, thus, this study cannot justify a single a best preference for which a data transformation can be used in all cases. Nevertheless, the assessment accomplished in this study provided an insight that COIA is sensitive to weights of species variables which varies whenever different data transformations are carried out in the ordination process. However, considering the characteristics of each data transformations and the nature of the data, the researcher strongly recommends the use of square root transformation.