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:Virtual BSU Seminar: 'Assumption-lean inference fo
r generalised linear model parameters' - Prof Stij
n Vansteelandt\, Ghent University
DTSTART;TZID=Europe/London:20200924T140000
DTEND;TZID=Europe/London:20200924T150000
UID:TALK150706AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/150706
DESCRIPTION:Inference for the parameters indexing generalised
linear models is routinely based on the assumption
that the model is correct and a priori specified.
This is unsatisfactory because the chosen model i
s usually the result of a data-adaptive model sele
ction process\, which induces bias and excess unce
rtainty that is not usually acknowledged\; moreove
r\, the assumptions encoded in the resulting model
rarely represent some a priori known\, ground tru
th. Standard inferences may therefore lead to bias
in effect estimates\, and may moreover fail to gi
ve a pure reflection of the information that is co
ntained in the data. Inspired by developments on a
ssumption-free inference for so-called projection
parameters\, we here propose nonparametric definit
ions of main effect estimands and effect modificat
ion estimands. These reduce to standard main effec
t and effect modification parameters in generalise
d linear models when these models are correctly sp
ecified\, but continue to capture the primary (con
ditional) association between two variables\, or t
he degree to which two variables interact (in a st
atistical sense) in their effect on outcome\, even
when these models are misspecified. We achieve an
assumption-lean inference for these estimands by
deriving their influence curve under the nonparame
tric model and invoking flexible data-adaptive (e.
g.\, machine learning) procedures. This talk aims
to be broadly accessible\, focussing on concepts m
ore than technicalities.
LOCATION:Virtual Seminar
CONTACT:Alison Quenault
END:VEVENT
END:VCALENDAR