Cluster analysis can not only cluster observations/cases into several groups but also cluster variables into several categories. Huge amount of categorical data is coming from different areas of research, both social and nature sciences. Therefore, there is a need to choose an appropriate method to analyze categorical data. This thesis starts with different clustering methods to cluster observations into several groups using categorical data. Three commonly used methods are compared and latent class analysis is recommended for categorical data analysis....[ Read more ]

Cluster analysis can not only cluster observations/cases into several groups but also cluster variables into several categories. Huge amount of categorical data is coming from different areas of research, both social and nature sciences. Therefore, there is a need to choose an appropriate method to analyze categorical data. This thesis starts with different clustering methods to cluster observations into several groups using categorical data. Three commonly used methods are compared and latent class analysis is recommended for categorical data analysis.

Dimensionality is an application of clustering variables/items into several categories/subscales in education. In this thesis, we develop an alternative approach to study the dimensionality. We are aware of no DETECT or PolyDETECT program available as part of the commonly used statistical packages such as SAS or Matlab. We thus develop a Matlab program to analyze the dimensional structure of educational test based on cluster analysis.