BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY: Latent class models for the causal effect of a treatment - Antoni
 o Forcina (University of Perugia)
DTSTART:20080509T130000Z
DTEND:20080509T140000Z
UID:TALK11785@talks.cam.ac.uk
CONTACT:8047
DESCRIPTION:In randomized experiments involving humans\, there is often ev
 idence that subjects who tend not to comply are also more likely to benefi
 t less (or perhaps more) from the treatment. Such situations may be modell
 ed by a directed acyclic graph involving a latent variable which represent
 s unobserved individual attitudes related to compliance and response. \nIn
  this talk a probabilistic approach free of counterfactuals will be presen
 ted. Though this model is not identifiable\, the conditions under which a 
 suitable average causal effect (across latent classes) is equal to the ins
 trumental variable estimand will be discussed. These conditions are the pr
 obabilistic analog of the conditions derived in the counterfactual literat
 ure (Angrist\, Imbens and Rubin\, Jasa\, 1996\; Hermàn and Robins\, Epide
 miology\, 2006). \n\nWhen additional information\, like individual covaria
 tes\, is available\, probabilistic models involving causal effects within 
 latent classes may become identifiable\; a technique for checking model id
 entifiability will be outlined together with numerical methods for computi
 ng maximum likelihood estimates of the parameters of interest. However\, e
 ven with additional information\, the price for model identifiabilility is
  a suitable set of modelling assumption. A few examples will be used as an
  illustration. \n\n\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR