www.uhasselt.be
DSpace

Document Server@UHasselt >
Research >
Research publications >

Please use this identifier to cite or link to this item: http://hdl.handle.net/1942/9582

Title: Identification of Salmonella high risk pig-herds in Belgium by using semiparametric quantile regression
Authors: BOLLAERTS, Kaatje
AERTS, Marc
RIBBENS, S.
VAN DER STEDE, Y.
BOONE, I.
Mintiens, K.
Issue Date: 2008
Publisher: BLACKWELL PUBLISHING
Citation: JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY, 171. p. 449-464
Abstract: Consumption of pork that is contaminated with Salmonella is an important source of human salmonellosis world wide. To control and prevent salmonellosis, Belgian pig-herds with high Salmonella infection burden are encouraged to take part in a control programme supporting the implementation of control measures. The Belgian government decided that only the 10% of pig-herds with the highest Salmonella infection burden (denoted high risk herds) can participate. To identify these herds, serological data reported as sample-to-positive ratios (SP-ratios) are collected. However, SP-ratios have an extremely skewed distribution and are heavily subject to confounding seasonal and animal age effects. Therefore, we propose to identify the 10% high risk herds by using semiparametric quantile regression with P-splines. In particular, quantile curves of animal SP-ratios are estimated as a function of sampling time and animal age. Then, pigs are classified into low and high risk animals with high risk animals having an SP-ratio that is larger than the corresponding estimated upper quantile. Finally, for each herd, the number of high risk animals is calculated as well as the beta-binomial p-value reflecting the hypothesis that the Salmonella infection burden is higher in that herd compared with the other herds. The 10% pig-herds with the lowest p-values are then identified as high risk herds. In addition, since high risk herds are supported to implement control measures, a risk factor analysis is conducted by using binomial generalized linear mixed models to investigate factors that are associated with decreased or increased Salmonella infection burden. Finally, since the choice of a specific upper quantile is to a certain extent arbitrary, a sensitivity analysis is conducted comparing different choices of upper quantiles.
Notes: [Bollaerts, Kaatje; Aerts, Marc] Hasselt Univ, Ctr Stat, B-3590 Diepenbeek, Belgium. [Ribbens, Stefaan] Univ Ghent, Merelbeke, Belgium. [Van der Stede, Yves; Boone, Ides; Mintiens, Koen] Vet & Agrochem Res Ctr, Brussels, Belgium.
URI: http://hdl.handle.net/1942/9582
DOI: 10.1111/j.1467-985X.2007.00525.x
ISI #: 000253890800009
ISSN: 0964-1998
Category: A1
Type: Journal Contribution
Validation: ecoom, 2009
Appears in Collections: Research publications

Files in This Item:

Description SizeFormat
Postprint6.95 MBAdobe PDF

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.