Readme file
SERIES C
 Applied Statistics
A Bayesian ordinal logistic regression model to correct for interobserver measurement error in a geographical oral health study, by S. M. Mwalili et al
Journal of the Royal Statistical Society,
Series C, Applied Statistics, Volume
54 (2005) part 1, 77-93
DESCRIPTION OF THE DATASET
A) THE SIGNAL TANDMOBIEL®DATASET
This is the data of the first year of the longitudinal Signal Tandmobiel® study (1996-2001) on Oral Health. There were 4468 children
involved in the study, and 81 variables were recorded in total. In the paper, the dataset is referred to as the main or cross-sectional
dental study.
Response: dmft, which is the sum of the number of decayed (d), missing due to caries (m) and filled (f) teeth in primary denition. Its values range from 0 to 20. The response is split according into:-
1 if dmft =0
2 if dmft =1
3 if 1<dmft<=4
4 if 4<dmft<=20
Covariates:the following covariates were of interest:-
- xcen,ycen: the tandardized form of the (x,y) coordinate of the municipality of the school to which the child belongs.
- age : the age of the children in years
- gender : the gender of the children, coded 0(boy) and 1 (girl)
- examiner : coded 1:16 for the 16 examiners, and 17 for the old standard
- schocode : an index of the schools to which the child belongs.
B) CALIBRATION DATASET
This is a validation data set containing information about the score for each of the 16 examiners against the gold standard. These data are 4 by 4 matrices for each examiner, with the columns (1-4) representing the gold standard score and the rows (1-4) corresponding to examiner's score.
DESCRIPTION OF THE PROGRAM
A WinBugs program was written to perform the analysis of the Bayesian ordinal logistic regression model to correct for Inter-observer
Measurement Error. The program is split into two parts:
- Main Data Model : Sampling a Markov chain pertaining to the regression parameters using correction term sampled from the calibration model
- Calibration Data Model: A Markov chain Sampling the correction terms using the calibration data
NOTE: these two two Markov chains are processed in sequential manner, by the use of the 'cut' option in WinBUGS.
Emmanuel Lesaffre
Biostatistical Centre
Katholieke Universiteit Leuven
Kapucynenvoer 35
B-300 Leuven
Belgium
E-mail: emmanuel.lesaffre@med.kuleuven.ac.be
- Datasets (Program_and_Data.txt, size - 30KB)
|
