A probabilistic model for co-occurrence analysis in bibliometrics.

Area of Science:

• Bibliometrics
• Information Science
• Computational Biology

Background:

• Co-occurrence analysis of Medical Subject Heading (MeSH) terms is vital in bibliometrics.
• Filtering low-frequency terms from co-occurrence matrices is crucial for interpretability but lacks standardized methods.
• Existing methods often struggle to objectively remove noise from co-occurrence data.

Purpose of the Study:

• To propose a novel probabilistic model for co-occurrence analysis.
• To provide a statistically-grounded method for determining critical thresholds in co-occurrence matrices.
• To reduce the dimensionality of co-occurrence matrices by removing random co-occurrences.

Main Methods:

• Development of a probabilistic model to assess the randomness of item co-occurrence.
• Application of statistical inferences to identify and filter insignificant co-occurring terms.
• Validation through conceptual framework, simulation studies, and practical bibliometric applications.

Main Results:

• The proposed probabilistic model effectively identifies random co-occurrences in MeSH term matrices.
• The model facilitates statistically driven dimensionality reduction of co-occurrence matrices.
• Demonstrated utility in enhancing the interpretability of bibliometric analyses.

Conclusions:

• The probabilistic model offers a robust solution for noise reduction in MeSH co-occurrence analysis.
• This method improves the reliability and interpretability of bibliometric findings.
• The approach provides a statistically sound basis for filtering co-occurrence data.