Effects of Dependence in High-Dimensional Multiple Testing Problems
Author Information
Author(s): Kim Kyung In, van de Wiel Mark A
Primary Institution: Eindhoven University of Technology
Hypothesis
How do dependence structures among variables affect the performance of False Discovery Rate (FDR) control procedures in high-dimensional data?
Conclusion
The adaptive Benjamini-Hochberg procedure is the most robust method for controlling the False Discovery Rate under dependence conditions.
Supporting Evidence
- The study shows that methods like SAM and the q-value do not adequately control the FDR under dependence.
- The adaptive Benjamini-Hochberg procedure is found to be the most robust method.
- The simulation results indicate that the estimates of the number of true null hypotheses vary under different dependence conditions.
Takeaway
When scientists test many hypotheses at once, they can make mistakes. This study shows how to better control those mistakes when the tests are related to each other.
Methodology
The study used simulations to compare the performance of various FDR procedures under different dependence structures among variables.
Potential Biases
Potential biases may arise from the assumptions made about the distribution of p-values and the nature of dependencies.
Limitations
The study primarily focuses on specific correlation structures and may not generalize to all types of dependencies.
Statistical Information
P-Value
0.1
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website