BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Data compression with statistical guarantees - Sylvia Richardson (
University of Cambridge)
DTSTART:20170703T123000Z
DTEND:20170703T131500Z
UID:TALK73129@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:Joint talk with Daniel Ahfock (MRC Biostatistics Unit @ Univer
sity of Cambridge)
The talk is concerned with translating r
ecent ideas from computer science on probabilistic data-compression
techniques into a statistical framework that can be &lsquo\;safely&rsquo\
; applied for speeding linear regression analyses for very larges sample s
izes in bio-medicine.
\;Our motivation is to facilitate
the use of multivariate regression and model exploration in tall data set
s\, so that\, for example\, genetic association analyses carried out on hu
ndreds of thousands of subjects can investigate multivariate effects for a
set of explanatory features\, rather than be restricted to one feature at
a time associations for computational feasibility.
Among th
e many approaches to dealing with tall data\, probabilistic data compressi
on techniques using random linear mapping\, developed in the comput
er science community\, so called sketching\, are particularly suita
ble for linear regression problems. In the first part of the talk\, we wil
l present a hierarchical representation of sketching\, which allows derivi
ng statistical properties (distributional) of different sketching algorith
ms. In particular\, we will discuss how the signal to noise ratio in the o
riginal data set is important for the choice of sketching algorithm. In th
e second part of the talk\, we will further refine some of the approximati
on guarantees and consider iterative sketches. The talk will be illustrate
d on a genetic analysis of the link between a blood cell trait and the HLA
region involving a sample of 130\,000 people.
http://arxi
v.org/abs/1706.03665
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR