Improving False Discovery Rate Control in Correlated Data
Author Information
Author(s): Lu Xin, Perkins David L
Primary Institution: UC San Diego
Hypothesis
Can a re-sampling strategy improve the estimation of the number of null hypotheses in FDR control under strong correlation structures?
Conclusion
The proposed re-sampling methods provide stronger control of the False Discovery Rate with only a minor sacrifice in power compared to traditional methods.
Supporting Evidence
- The re-sampling methods reduced the variation in estimating the proportion of true null hypotheses.
- The methods outperformed traditional FDR control methods in terms of false discovery rate.
- Simulation studies showed that the proposed methods maintained a high level of power.
Takeaway
When scientists test many genes at once, they need to avoid mistakes. This study shows a new way to do that better, especially when genes are related.
Methodology
The study developed two re-sampling methods to estimate the proportion of true null hypotheses in datasets with strong correlations.
Potential Biases
The estimation may be biased but conservative, leading to fewer rejections than other methods.
Limitations
The methods may not be directly applicable to more complex problems without sufficient replicates.
Participant Demographics
The study used simulated datasets based on actual microarray data from breast cancer cases.
Statistical Information
P-Value
0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website