New Method for Evaluating Cluster Stability
Author Information
Author(s): Gana Dresen, Boes Tanja, Huesing Johannes, Neuhaeuser Markus, Joeckel Karl-Heinz
Primary Institution: Institut für Medizinische Informatik, Biometrie und Epidemiologie, Universitaetsklinikum Essen, Germany
Hypothesis
Can a new resampling method based on continuous weights improve the assessment of cluster stability in hierarchical clustering compared to conventional bootstrapping?
Conclusion
Continuous weights produce at least the same results as conventional bootstrapping and in some cases surpass it.
Supporting Evidence
- Continuous weights retain all observations in the resampled dataset, unlike bootstrapping.
- In small datasets, continuous weights provide more informative results than bootstrapping.
- The method was validated using real datasets and simulation studies.
Takeaway
This study shows a new way to check if groups of data are real or just random by using special weights, which helps keep all the information from the data.
Methodology
The study compared a new resampling method using continuous weights to conventional bootstrapping on real and simulated datasets.
Limitations
The method's performance may vary with the number of features and the nature of the datasets.
Participant Demographics
The study involved datasets from 20 patients with uveal melanoma and 22 primates.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website