False Discovery Rate-Controlled Gene Selection under Variance Heterogeneity and Correlated Expression Measurements

SCHIMEK, Michael G. a Tomáš PAVLÍK. False Discovery Rate-Controlled Gene Selection under Variance Heterogeneity and Correlated Expression Measurements. In International Biometric Conference. 2008. ISBN 978-0-9821919-0-3.

Základní údaje

• Originální název: False Discovery Rate-Controlled Gene Selection under Variance Heterogeneity and Correlated Expression Measurements
• Název anglicky: False Discovery Rate-Controlled Gene Selection under Variance Heterogeneity and Correlated Expression Measurements
• Autoři: SCHIMEK, Michael G. a Tomáš PAVLÍK.
• Vydání: International Biometric Conference, 2008.

Další údaje

• Typ výsledku: Vyžádané přednášky
• Utajení: není předmětem státního či obchodního tajemství
• Organizační jednotka: Lékařská fakulta
• ISBN: 978-0-9821919-0-3

Anotace

We demonstrate that the comparative findings obtained for the above FDR procedures do not depend as much on the correlation structure than on the variance heterogeneity of the involved gene expression values. Both aspects are contributing to the complexity of the data model. In low complexity models the "correcting" constant for variance control can be quite useful, assuming it is well chosen. Finally we provide guidelines for the practical use of the available SAM techniques. Last but not least, we compare the SAM results to those of a FDR-free empirical Bayes procedure due to Johnstone and Silverman (2004).