Normal view MARC view ISBD view

Clustering-based procedures for identifying multiple outliers in univariate circular biological data / Nur Syahirah Zulkipli

By:

Nur Syahirah Zulkipli [author.]

Material type: Text

text

Media type:

unmediated

Carrier type:

volume

ISBN:

THE0010820(Local)

Subject(s):

Dissertation note: Thesis (Doctor of Philosophy) -- Universiti Malaysia Pahang - 2026 Abstract: Circular data analysis has gained increasing attention in biological and biomedical research, particularly in improving outlier detection methods for univariate circular data. However, existing methods such as row deletion methods remain limited in detecting multiple outliers in univariate circular data. Although clustering-based procedures have been applied to bivariate circular data, their extension to univariate circular settings remains challenging due to structural differences. Hence, this study proposes a clusteringbased outlier detection (COD) procedure for univariate circular biological data using agglomerative hierarchical clustering algorithms such as single-linkage (SL), completelinkage (CL), and average-linkage (AL), combined with Satari/Di and Chang circular similarity measures to address the limitations of existing methods in detecting multiple outliers. The proposed COD procedure was evaluated through simulation studies based on the von Mises distribution under three outlier scenarios representing different contamination conditions. The findings indicate that the performance of the COD procedure is influenced by the concentration parameter, percentage of outliers and sample size. Overall, the CL-Satari/Di algorithm demonstrated strong outlier detection capability with minimal masking effects, while the SL-Satari/Di algorithm showed greater resistance to swamping effects and provided more stable performance across historical datasets. The proposed COD procedure was further validated using historical univariate circular biological datasets and syndromic craniosynostosis data in Malaysia, where it successfully identified both single and multiple outliers. Moreover, compared with existing A statistic row deletion method, the proposed COD procedure demonstrated improved capability in detecting multiple outliers while reducing masking and swamping effect. This study contributes to circular statistics methodology by extending clusteringbased outlier detection techniques to circular data, providing insight into the behaviour of masking and swamping effects in circular spaces, and also improving the understanding of contamination and cluster separability in univariate circular biological data.

Tags from this library: No tags from this library for this title. Log in to add tags.

Average rating: 0.0 (0 votes)

Holdings ( 2 )
Title notes ( 4 )

Holdings
Item type	Current library	Collection	Call number	Copy number	Status	Date due	Barcode
Thesis	UMPLIB GAMBANG	Reference	PSM .S93 2026 r Thesis (Browse shelf(Opens below))	1	In Transit		T000005624
Accompanying Material	UMPLIB GAMBANG	Reference	CD14051 (Browse shelf(Opens below))	1	In Transit		T000005625

Center for Mathematical Sciences

Thesis (Doctor of Philosophy) -- Universiti Malaysia Pahang - 2026

Includes bibliographical reference

Circular data analysis has gained increasing attention in biological and biomedical research, particularly in improving outlier detection methods for univariate circular data. However, existing methods such as row deletion methods remain limited in detecting multiple outliers in univariate circular data. Although clustering-based procedures have been applied to bivariate circular data, their extension to univariate circular settings remains challenging due to structural differences. Hence, this study proposes a clusteringbased outlier detection (COD) procedure for univariate circular biological data using agglomerative hierarchical clustering algorithms such as single-linkage (SL), completelinkage (CL), and average-linkage (AL), combined with Satari/Di and Chang circular similarity measures to address the limitations of existing methods in detecting multiple outliers. The proposed COD procedure was evaluated through simulation studies based on the von Mises distribution under three outlier scenarios representing different contamination conditions. The findings indicate that the performance of the COD procedure is influenced by the concentration parameter, percentage of outliers and sample size. Overall, the CL-Satari/Di algorithm demonstrated strong outlier detection capability with minimal masking effects, while the SL-Satari/Di algorithm showed greater resistance to swamping effects and provided more stable performance across historical datasets. The proposed COD procedure was further validated using historical univariate circular biological datasets and syndromic craniosynostosis data in Malaysia, where it successfully identified both single and multiple outliers. Moreover, compared with existing A statistic row deletion method, the proposed COD procedure demonstrated improved capability in detecting multiple outliers while reducing masking and swamping effect. This study contributes to circular statistics methodology by extending clusteringbased outlier detection techniques to circular data, providing insight into the behaviour of masking and swamping effects in circular spaces, and also improving the understanding of contamination and cluster separability in univariate circular biological data.