BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Statistics
SUMMARY:Two-sample testing of high-dimensional linear regr
ession coefficients via complementary sketching -
Tengyao Wang (LSE)
DTSTART;TZID=Europe/London:20220121T140000
DTEND;TZID=Europe/London:20220121T150000
UID:TALK168029AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/168029
DESCRIPTION:We introduce a new methodology for two-sample test
ing of high-dimensional linear regression coeffici
ents without assuming that those coefficients are
individually estimable. The procedure works by fir
st projecting the matrices of covariates and respo
nse vectors along directions that are complementar
y in sign in a subset of the coordinates\, a proce
ss which we call 'complementary sketching'. The re
sulting projected covariates and responses are agg
regated to form two test statistics. We show that
our procedure has essentially optimal asymptotic p
ower under Gaussian designs with a general class o
f design covariance matrices when the difference b
etween the two regression coefficients is sparse a
nd dense respectively. Simulations confirm that ou
r methods perform well in a broad class of setting
s.
LOCATION:MR12\, Centre for Mathematical Sciences
CONTACT:Qingyuan Zhao
END:VEVENT
END:VCALENDAR