@mikelove @jkpritch @tkyzeng @stephaniehicks @will_townes @keegankorthauer @siminaboca @jtleek Further related work: https://t.co/I2FR91IkFM uses Boosting along with AdaPT (@Stat_Ron), the method of @siminaboca, @jtleek (https://t.co/5wUSYiQ5sH) can be used along with e.g., boosting to estimate the prior probability of being null as a function of high-dim covariates.
@LuciaScience @mikelove @stephaniehicks @daniela_witten @drisso1893 @pkkimes @keegankorthauer I think we noted this in our paper (https://t.co/Z2uL1cwXgz with @jtleek, which is one of the methods in @stephaniehicks's review mentioned above), which extends Storey's approach to include covariates in the estimation of pi_0 (so has same caveats w/ nr of tests.)
@jschwart37 @chr1swallace @GReales7 @MRC_BSU (1/n) We now additionally compare Flexible cFDR to IHW(https://t.co/lwthEQssGu…) and Boca and Leek’s FDR regression (BL; https://t.co/XYcynhkDKS) (and GenoWAP https://t.co/KWTtGPTTqT…) and FINDOR (https://t.co/iDxLsssm7s…) as in the previous manuscript version).
Our methods paper is at https://t.co/Z2uL1cfmp1 and the development version (or "pre-development" since it's not in devel/bioc yet) of our package is at https://t.co/J0xFhewPlH. #bioc2020
@jtleek @deaneckles @keegankorthauer @stephaniehicks In this paper with @jtleek we also note that FDR control is better with 10,000 vs 1,000 tests https://t.co/Z2uL1cfmp1
A direct approach to estimating false discovery rates conditional on covariates https://t.co/9R0PEyNx2o #bioinformatics #DataScience https://t.co/6QzE4WaVQB