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:MRC Biostatistics Unit Seminars
SUMMARY:Sensitivity of parameter estimates of marginal and
random-effects models to missing data - Rumana Om
ar\, Department of Statistical Science\, UCL
DTSTART;TZID=Europe/London:20110315T143000
DTEND;TZID=Europe/London:20110315T153000
UID:TALK28338AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/28338
DESCRIPTION:Random effects (RE) models and marginal models bas
ed on generalised estimating equations (GEE) are f
requently used to analyse longitudinal repeated me
asurements health studies where subject dropout is
common. RE models require MAR assumption. Because
marginal models are not based on likelihoods\, th
ey require data to be MCAR. When the data are Gaus
sian\, the GEEs reduce to score equations and prov
ided the correct correlation structure is applied
the two types of models are equivalent and margina
l models should then be robust to MAR. However\, e
quivalent marginal and RE models for Gaussian data
may not necessarily produce identical parameter e
stimates due to missing data as GEEs may not reduc
e to score equations in that situation\, even in p
resence of MCAR. For binary data the marginal mode
ls require MCAR assumption. By definition neither
RE or marginal models are robust to MNAR.\n \nI
n practice\, the extent to which missing observati
ons cause bias to the parameter estimates of these
models and affect their clinical and statistical
significance is not clear. Limited simulation stud
ies have been conducted. However\, it is not clear
what proportion of missingness leads to substanti
al bias or how sensitivity to missing data compare
s between cluster-level and cluster-varying covari
ates. It is not known to what extent the marginal
model is robust to misspecification of the working
correlation matrix in presence of missing data an
d whether the strength of the intracluster correla
tion coefficient affects the bias in parameter est
imates caused by MNAR. The aim here is to explore
the effects of dropout on parameter estimates of R
E and marginal models for repeated measurements da
ta for both Gaussian and binary outcome using simu
lation studies in order to make some practical re
commendations regarding analyses. \n
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Publ
ic Health\, University Forvie Site\, Robinson Way\
, Cambridge
CONTACT:Michael Sweeting
END:VEVENT
END:VCALENDAR