BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Mixed effects models with covariates perturbed for SDC - Silvia Po
 lettini (Università degli Studi di Roma La Sapienza)
DTSTART:20161209T141500Z
DTEND:20161209T150000Z
UID:TALK69411@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Co-author: Serena Arima (Sapienza Universit&agrave\; di 
  Roma)<br><br>We focus on mixed effects with data subject to PRAM. An inst
 ance of this is a&nbsp\;small area model.&nbsp\;&nbsp\;We assume&nbsp\;tha
 t categorical covariates have been perturbed by Post Randomization\,<br>wh
 ereas the level identifier is not perturbed. We also&nbsp\;assume&nbsp\;th
 at a continuous response&nbsp\;is available\, &nbsp\;and consider a nested
  linear regression model:<br>$$<br>y_{ij}= X_{ij}^{&#39\;}\\beta +v_{i}+e_
 {ij}\, &nbsp\; &nbsp\; j=1\,...\,n_{i}\; \\\,\\\,i=1\,...\,m<br>$$<br>wher
 e<br>$v_{i}\\iid N(0\,\\sigma^{2}_{v})$ (model error)\;$e_{i}\\iid<br>N(\\
 mu\,\\sigma^{2}_{e})$ (design error).<br><br>We resort to a measurement er
 ror model and define a unit-level small&nbsp\;area model accounting for me
 asurement error &nbsp\;in &nbsp\; discrete covariates.<br>PRAM is defined 
 in terms of &nbsp\;a transition matrix $P$ modeling the&nbsp\;changes in c
 ategories\;&nbsp\;we consider both the case of known $P$\, and the case wh
 en &nbsp\;$P$ is<br>unknown and is estimated from the data.<br><br>A small
  simulation study is conducted to assess the effectiveness of&nbsp\;the pr
 oposed Bayesian measurement error model in estimating the model<br>paramet
 ers and to investigate the protection provided by PRAM in this&nbsp\;conte
 xt.</span>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR