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:Variable elimination and graph reduction: towards 
 an efficient g-formula for causal DAGs - Richard G
 uo (University of Cambridge)
DTSTART;TZID=Europe/London:20211119T140000
DTEND;TZID=Europe/London:20211119T150000
UID:TALK162136AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/162136
DESCRIPTION:Consider a study where the causal structure is kno
 wn and described by a directed acyclic graph (DAG)
 . A causal quantity of interest\, say a counterfac
 tual mean\, can often be expressed as a functional
  of the observed distribution given by the g-formu
 la (also known as the "truncated factorization"). 
 The g-formula\, which can be written down from the
  graph\, usually takes the form of an integral inv
 olving conditional expectations of the variables i
 n the graph.\n\nNaturally\, to estimate the causal
  quantity efficiently\, one can use a plugin estim
 ator of the g-formula\, where every conditional ex
 pectation is replaced by its MLE. However\, we fin
 d that asymptotically not every variable appearing
  in the g-formula carries information for estimati
 on. In fact\, the causal quantity can often be est
 imated with an "efficient" g-formula that drops th
 e redundant variables such that the cost of measur
 ing these variables can be saved.\n\nWe present a 
 graphical procedure towards this goal. First\, we 
 identify a set of graphical conditions that are ne
 cessary and sufficient for eliminating redundant v
 ariables. Second\, we construct a reduced DAG on t
 he non-redundant variables only\, from which the "
 efficient" g-formula can be derived. The reduced D
 AG is transformed from the original DAG through a 
 set of "moves"\, traversing both within and betwee
 n Markov equivalence classes\, which nonetheless p
 reserve the semiparametric efficiency bound for es
 timating the causal quantity.
LOCATION:MR12\, Centre for Mathematical Sciences
CONTACT:Qingyuan Zhao
END:VEVENT
END:VCALENDAR