BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Parameterizing and Simulating from Causal Models - Robin Evans (Un
 iversity of Oxford)
DTSTART:20230217T140000Z
DTEND:20230217T150000Z
UID:TALK194902@talks.cam.ac.uk
CONTACT:Qingyuan Zhao
DESCRIPTION:Many statistical problems in causal inference involve a probab
 ility distribution other than the one from which data are actually observe
 d\; as an additional complication\, the object of interest is often a marg
 inal quantity of this other probability distribution. This creates many pr
 actical complications for statistical inference\, even where the problem i
 s non-parametrically identified. In particular\, it is difficult to perfor
 m likelihood-based inference\, or even to simulate from the model in a gen
 eral way. \n\nWe introduce the 'frugal parameterization'\, which places th
 e causal effect of interest at its centre\, and then build the rest of the
  model around it. We do this in a way that provides a recipe for construct
 ing a regular\, non-redundant parameterization using causal quantities of 
 interest. In the case of discrete variables we can use odds ratios to comp
 lete the parameterization\, while in the continuous case copulas are the n
 atural choice\; other possibilities are also discussed. \n\nOur methods al
 low us to construct and simulate from models with parametrically specified
  causal distributions\, and fit them using likelihood-based methods\, incl
 uding fully Bayesian approaches. Our proposal includes parameterizations f
 or the average causal effect and effect of treatment on the treated\, as w
 ell as other common quantities of interest.  \n \nI will also discuss some
  other applications of the frugal parameterization\, including to survival
  analysis\, parameterizing nested Markov models\, and ‘Many Data’: com
 bining randomized and observational datasets in a single parametric model.
 \n \nThis is joint work with Vanessa Didelez (University of Bremen and Lei
 bniz Institute for Prevention Research and Epidemiology - BIPS).\n \nRefer
 ence\nEvans\, R.J. and Didelez\, V.  Parameterizing and Simulating from Ca
 usal Models\, arXiv preprint:2109.03694\, 2021.
LOCATION:MR12\, Centre for Mathematical Sciences
END:VEVENT
END:VCALENDAR