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:Nonlinear Shrinkage Estimation in Quadratic Infere
nce Function Analysis for Correlated Data - Cliffo
rd Lam (London School of Economics)
DTSTART;TZID=Europe/London:20180215T110000
DTEND;TZID=Europe/London:20180215T120000
UID:TALK101095AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/101095
DESCRIPTION:Quadratic inference function (QIF) analysis is mor
e efficient than the generalized estimating equati
ons (GEE) approach when the working covariance mat
rices for the data are misspecified. Since QIF nat
urally requires a weighting matrix which is the in
verse of a sample covariance matrix of non-identic
ally distributed data\, finite sample performance
can be greatly affected when the number of indepen
dent data points is not large enough\, which is us
ually the case in cluster randomized trials or man
y longitudinal studies. While nonlinear shrinkage
is very successful in regularizing the extreme eig
envalues of a sample covariance matrix\, the metho
d is only restricted to independent and identicall
y distributed data. We propose a novel nonlinear s
hrinkage approach for a sample covariance matrix o
f non-identically distributed data\, which improve
s finite sample performance of QIF\, and gives roo
m for increasing the potential number of working c
orrelation structures for even better performance.
Together with a nonlinearly shrunk weighting matr
ix\, we derive the asymptotic normality of the par
ameter estimators\, and give an estimator for the
asymptotic covariance matrix. We demonstrate the p
erformance of the proposed method through simulati
on experiments and a real data analysis.

LOCATION:Seminar Room 2\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR