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:Spatial causal inference in the presence of unmeas
ured confounding and interference - Georgia Padado
georgou\, University of Florida
DTSTART;TZID=Europe/London:20240426T140000
DTEND;TZID=Europe/London:20240426T150000
UID:TALK213460AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/213460
DESCRIPTION:In this talk\, I aim to bridge the divide between
causal inference and spatial statistics\, by prese
nting novel insights for causal inference in spati
al data analysis and establishing how tools from s
patial statistics can be used to draw causal infer
ences. I will introduce spatial causal graphs to h
ighlight that spatial confounding and interference
can be entangled\, in that investigating the pres
ence of one can lead to wrongful conclusions in th
e presence of the other. Moreover\, I will illustr
ate that spatial dependence in the exposure variab
le can render standard analyses invalid\, which ca
n lead to erroneous conclusions. To remedy these i
ssues\, we propose a Bayesian parametric approach
based on tools commonly-used in spatial statistics
. This approach simultaneously accounts for interf
erence and mitigates bias resulting from local and
neighbourhood unmeasured spatial confounding. Fro
m a Bayesian perspective\, we show that incorporat
ing an exposure model is necessary\, and we theore
tically prove that all model parameters are identi
fiable\, even in the presence of unmeasured confou
nding. We study the impact of sulfur dioxide emiss
ions from power plants on cardiovascular mortality
.
LOCATION:MR12\, Centre for Mathematical Sciences
CONTACT:Dr Sergio Bacallado
END:VEVENT
END:VCALENDAR