Evaluating Variable Selection in PCA Visualization
Author Information
Author(s): Fontes Magnus, Soneson Charlotte
Primary Institution: Centre for Mathematical Sciences, Lund University
Hypothesis
There exists no objective method for determining the optimal inclusion criterion in the context of visualization.
Conclusion
The projection score provides an easily interpretable and universally applicable measure of the informativeness of a variable subset with respect to visualization by PCA.
Supporting Evidence
- The projection score can be applied universally to find suitable inclusion criteria for variable filtering.
- The study demonstrates the effectiveness of the projection score in improving PCA visualizations.
- The projection score traces out a smooth curve indicating a maximally informative subset.
- The study validates the projection score with synthetic examples.
Takeaway
This study introduces a new way to pick the best variables for making sense of complex data, helping scientists see patterns more clearly.
Methodology
The study applies the projection score to find optimal variable subsets for different filtering methods in microarray and synthetic data sets.
Potential Biases
The use of ad-hoc criteria for variable selection may introduce bias.
Limitations
The study does not address the potential biases introduced by the variable selection process.
Participant Demographics
The study includes data from 56 subjects with various cancer types.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website