Classification to ordinal categories using a search partition methodology with an application in diabetes screening

Area of Science:

• Statistics
• Biostatistics
• Machine Learning

Background:

• Ordinal data classification presents unique challenges.
• Existing methods may not fully capture the ordered nature of categories.
• The Search Partition Analysis (SPAN) approach offers a potential solution.

Purpose of the Study:

• To introduce and evaluate a novel SPAN-based method for ordinal category classification.
• To apply the method to glucose tolerance data for discriminating health states.
• To compare the performance of the new method against established techniques.

Main Methods:

• Repeatedly applying SPAN to binary outcomes derived from collapsed adjacent ordinal categories.
• Ensuring successive binary partitions are nested to form a final ordinal classification.
• Utilizing glucose tolerance data to classify individuals into diabetes, impaired glucose tolerance, and normal states.

Main Results:

• The proposed SPAN method successfully classifies ordinal categories.
• The approach demonstrates efficacy in discriminating between different glucose tolerance states.
• Performance comparison with ordinal logistic regression and classification trees is presented.

Conclusions:

• The nested binary partition SPAN approach provides a viable method for ordinal classification.
• This technique offers a valuable alternative for analyzing ordered categorical data, particularly in health-related contexts.
• Further research can explore its application in diverse fields with ordinal outcomes.