BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:RANK: Large-Scale Inference with Graphical Nonlinear Knockoffs - Y
 ingying Fan (University of Southern California)
DTSTART:20180116T110000Z
DTEND:20180116T114500Z
UID:TALK97633@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Co-authors: Emre Demirkaya		(University of Southern Cali
 fnornia)\, Gaorong Li		(Beijing University of Technology)\, Jinchi Lv		(Un
 iversity of Southern Califnornia)        <br></span><br>Power and reproduc
 ibility are key to enabling refined scientific discoveries in contemporary
  big data applications with general high-dimensional nonlinear models. In 
 this paper\, we provide theoretical foundations on the power and robustnes
 s for the model- free knockoffs procedure introduced recently in Cand`es\,
  Fan\, Janson and Lv (2016) in high-dimensional setting when the covariate
  distribution is characterized by Gaussian graphical model. We establish t
 hat under mild regularity conditions\, the power of the oracle knockoffs p
 rocedure with known covariate distribution in high-dimensional linear mode
 ls is asymptotically one as sample size goes to infinity. When moving away
  from the ideal case\, we suggest the modified model-free knockoffs method
  called graphical nonlinear knockoffs (RANK) to accommodate the unknown co
 variate distribution. We provide theoretical justifications on the robustn
 ess of our modified procedure by showing that the false discovery rate (FD
 R) is asymptoti cally controlled at the target level and the power is asym
 ptotically one with the estimated covariate distribution. To the best of o
 ur knowledge\, this is the first formal theoretical result on the power fo
 r the knock- offs procedure. Simulation results demonstrate that compared 
 to existing approaches\, our method performs competitively in both FDR con
 trol and power. A real data set is analyzed to further assess the performa
 nce of the suggested knockoffs procedure. <br><br>Related Links<ul><li><a 
 target="_blank" rel="nofollow" href="http://www-old.newton.ac.uk/cgi/http%
 3A%2F%2Fwww-bcf.usc.edu%2F~fanyingy%2Fpublications%2FRANK-FDLL17.pdf">http
 ://www-bcf.usc.edu/~fanyingy/publications/RANK-FDLL17.pdf</a></li></ul>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR