BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.177.160//NONSGML kigkonsult.se iCalcreator 2.18//
BEGIN:VEVENT
UID:20220518T191228EDT-1842Zirca6@132.216.177.160
DTSTAMP:20220518T231228Z
DESCRIPTION:Title: Nonparametric location-scale models for right- and inter
val-censored data with inference based on Laplace approximations.\n\nAbstr
act: A double additive model for the conditional mean and standard\n deviat
ion in location-scale models with a nonparametric error distribution is pr
oposed. The response is assumed continuous and possibly subject to right o
r interval-censoring. Nonparametric inference from censored data in locati
on-scale models has been studied by many authors\, but it generally focuse
s on the estimation of conditional location and can only deal with the est
imation of the smooth effects of a very limited number of covariates. Addi
tive models based on P-splines are preferred here for their excellent prop
erties and the possibility to handle a large number of additive terms (Eil
ers and Marx 2002). They are used to specify the joint effect of covariate
s on location and dispersion within the location-scale model. A nonparamet
ric error distribution with a smooth underlying hazard function and fixed
moments is assumed for the standardized error term. In the absence of righ
t-censoring\, a location-scale model with a small number of additive terms
and a quartile-constrained error density (instead of the hazard here) was
considered in Lambert (2013) to analyse interval-censored data\, with inf
erence relying on a numerically demanding MCMC algorithm. It is shown how
Laplace approximations to the conditional posterior of spline parameters c
an be combined to bring fast and reliable estimation of the linear and add
itive terms\, and provide a smooth estimate of the underlying error hazard
function under moment constraints. These approximations are the cornersto
nes in the derivation of the marginal posteriors for the penalty parameter
s and smoothness selection. The resulting estimation procedures are motiva
ted using Bayesian arguments and shown to own excellent frequentist proper
ties. They are extremely fast and can handle a large number of additive te
rms within a few seconds even with pure R code. The methodology is illustr
ated with the analysis of right- and interval-censored income data in a su
rvey.\n\n \n\n \n
DTSTART:20211130T153000Z
DTEND:20211130T163000Z
SUMMARY:Philippe LAMBERT (Université de Liège et Université catholique de L
ouvain\, Belgique)
URL:https://www.mcgill.ca/mathstat/channels/event/philippe-lambert-universi
te-de-liege-et-universite-catholique-de-louvain-belgique-335169
END:VEVENT
END:VCALENDAR