BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:"\;Better prediction by use of co-data: Adaptive group-regular ized ridge regression"\; - Dr Mark van de Wiel\, VU University Medical Center and VU university\, Amsterdam DTSTART:20160322T143000Z DTEND:20160322T153000Z UID:TALK65252@talks.cam.ac.uk CONTACT:Alison Quenault DESCRIPTION:For high-dimensional settings\, we show how one can use empiri cal Bayes (EB) principles to estimate penalties that may differ across gro ups of variables. These groups are predefined using co-data\, which is aux iliary information available on the variables (e.g. genomic annotation or external p-values). Due to the adaptive character of the penalties\, the g roup-wise penalties may improve predictions when the groups are indeed inf ormative\, while not deteriorating those when this is not the case. We pro vide an implementation in a classical logistic ridge regression setting. H owever\, we will also discuss extension of the framework to a Bayesian rid ge regression setting. The latter is particularly useful for obtaining cre dibility intervals on the predicted event probabilities. In particular a h ybrid EB-Full Bayes approach in combination with highest-probability densi ty intervals seem to have good coverage properties when the number of vari ables is not extremely large. Finally\, the potential for better variable selection\, either by post-hoc selection or by sparse regression\, will be shortly considered. Several real data examples will be discussed\, in par ticular on cancer diagnostics using a variety of molecular data types\, su ch as methylation\, RNAseq and microRNAs.\n LOCATION:Large Seminar Room\, 1st Floor\, Institute of Public Health\, Un iversity Forvie Site\, Robinson Way\, Cambridge END:VEVENT END:VCALENDAR