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:Isaac Newton Institute Seminar Series
SUMMARY:A Hybrid Block Bootstrap For Sample Quantiles Unde
r Weak Dependence - Alastair Young (Imperial Colle
ge London)
DTSTART;TZID=Europe/London:20180220T110000
DTEND;TZID=Europe/London:20180220T120000
UID:TALK101152AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/101152
DESCRIPTION:The subsampling bootstrap and the moving blo
cks bootstrap provide effective methods for nonpar
ametric inference with weakly dependent data. Both
are based on the notion of resampling (overlappin
g) blocks of successive observations from a data s
ample: in the former single blocks are sampled\, w
hile the latter splices together random blocks to
yield bootstrap series of the same length as the o
riginal data sample. Here we discuss a general the
ory for block bootstrap distribution estimation fo
r sample quantiles\, under mild strong mixing assu
mptions. A hybrid between subsampling and the movi
ng blocks bootstrap is shown to give theoretical b
enefits\, and startling improvements in accuracy i
n distribution estimation in important practical s
ettings. An intuitive procedure for empirical sele
ction of the optimal number of blocks and their le
ngth is proposed. The conclusion that bootstrap sa
mples should be of smaller size than the original
data sample has significant implications for compu
tational efficiency and scalability of bootstrap m
ethodologies in dependent data settings. This is j
oint work with Todd Kuffner and Stephen Lee and is
described at https://arxi
v.org/abs/1710.02537.
LOCATION:Seminar Room 2\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR