Your Cart

Your cart is empty

Home  >  Volume 50

1. THE EFFECT OF DIFFERENT SIMILARITY MEASURES ON TOPOLOGY OF TREES USING SIMULATED BINARY DATA by Ojurongbe T.A. and Schachtel G. Volume 50 (March, 2019 Issue)
Sale price: $0.00
Quantity:

Department of Mathematical Sciences, Osun State University, P.M.B 4494, Osogbo, Osun State.

Department of Biometry and Population Genetics, Justus Liebig University, Heinrich-Buff-Ring 26-32, D-35392 Giessen, Germany

Abstract

Binary data with varying parameters for rows and columns with distinct characteristics for grouping were generated and cluster analysis using the UPGMA (Unweighted Pair-Group Mean Arithmetic) method was carried out. It is widely believed that Dice and Jaccard measures usually give similar results with respect to cluster analysis. Bearing this in mind, the cluster analysis was based on these two measures with the aim of confirming whether this is really true. Therefore, the objective of this study was to investigate the impact of the underlying (chosen) similarity (dissimilarity) measure on the resulting classification from cluster analysis and the extent to which measure and classification do affect topology. Consensus Fork Index (CFI) was calculated to compare the trees generated from the cluster analysis for the measures that were chosen. Other analyses like Multi-dimensional scaling (MDS) and Principal Component Analysis (PCA) were carried out to confirm the similarities in the structures of the trees.  The results showed that CFI alone is not enough to determine the topology of a tree, likewise correlation. It was concluded that, in choosing a similarity coefficient for clustering analysis, Dice and Jaccard coefficients will always give similar results with room for a few exceptional cases and that classification goes a long way to affect topology.

Keywords: cluster analysis, topology, correlation, classification.

click here to download full abstract

Close
Loading...