Evaluating Jackknife and Bootstrap for Confidence Intervals in Clustering Agreement Measures
Author Information
Author(s): Severiano Ana, Carriço João A., Robinson D. Ashley, Ramirez Mário, Pinto Francisco R.
Primary Institution: Instituto de Microbiologia, Instituto de Medicina Molecular, Faculdade de Medicina de Lisboa, Lisboa, Portugal
Hypothesis
What are the appropriate methods for calculating confidence intervals for clustering agreement measures?
Conclusion
The jackknife method outperforms the bootstrap method for estimating confidence intervals for pairwise agreement measures.
Supporting Evidence
- The jackknife method showed superior coverage for confidence intervals across various parameters.
- Bootstrap methods resulted in lower coverage for low agreement values.
- Confidence intervals calculated by the jackknife were robust to changes in cluster number and size distribution.
Takeaway
This study looks at two ways to check how well different methods agree with each other, and finds that one method (jackknife) is better than the other (bootstrap) for getting accurate results.
Methodology
The study evaluated the performance of jackknife and bootstrap methods for estimating confidence intervals using simulated data.
Potential Biases
The bootstrap method may produce biased estimates when the sample agreement is low.
Limitations
The study highlights that small sample sizes can lead to unreliable estimates of confidence intervals.
Statistical Information
Confidence Interval
95%
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website