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:The data-driven (s\,S) policy: why you can have co
nfidence in censored demand data - Gah-Yi Vahn (Lo
ndon School of Business)
DTSTART;TZID=Europe/London:20160129T160000
DTEND;TZID=Europe/London:20160129T170000
UID:TALK63640AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/63640
DESCRIPTION:I revisit the classical dynamic inventory manageme
nt problem of Scarf (1959) from a distribution-fre
e\, data-driven perspective. I propose a nonparame
tric estimation procedure for the optimal (s\, S)
policy that yields an asymptotically optimal estim
ated policy and analytically derive confidence int
ervals around this policy. I also derive a confide
nce bound on the estimated total cost\, which\, in
the case of zero setup cost\, interestingly is di
rectly proportional to the size of the confidence
intervals of the estimated policy. I further consi
der having a portion of the data censored from pas
t ordering decisions. I show that the intuitive pr
ocedure of correcting for censoring in the demand
data directly yields an inconsistent estimate of t
he optimal policy. I then show how to correctly us
e the censored data to obtain consistent decisions
and derive confidence intervals for this policy.
Remarkably\, under some conditions\, ordering deci
sions estimated with partially censored data may b
e more precise than with fully uncensored data\, a
nd there exists an optimal amount of censored data
to minimise the mean square error (MSE).
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberfo
rce Road\, Cambridge.
CONTACT:Quentin Berthet
END:VEVENT
END:VCALENDAR