Readme file
SERIES C
 Applied Statistics
A Bayesian model for longitudinal count data with non-ignorable dropout, by N. A. Kaciroti, T. E. Raghunathan, M. A. Schork and N. M. Clark, pages 521–534;
Three separate files are used for the data analysis reported in this paper
1) WINBUGS Code: This file contains the WinBUGS Code
2) Data: This file contains the data
3) Initial Data: This file contains the initial values used to run the program in WinBUGS Code
DESCRIPTION OF THE FILES
1) WinBUGS Code file
This file contain the code used to produced the results reported in the paper. The parameters of interest, which are estimated and given in Table 1 and Table 2 are beta and gamma parameters along with the variance-covariance parameters for the random effects sigma_1, sigma_2, sigma_12 and sigma_g.
2) Data file
This file contains the data analyzed. The data consist of the following variables:
Physid: the physician id
Y0: the number of hospitalizations at baseline
Y1: the number of hospitalization at follow-up1
Y2: the number of hospitalization at follow-up2
Offset1: the number of months for the follow-up1
Offset2: the number of months for the follow-up2
Data are sorted by missing data patterns and by physid. Thus,
-1:N0 is the subset of subjects who have complete data
-N0+1:N1 is the subset of subjects who have data on Y0 and Y1 but not on Y2
-N1+1:N is the subset of subjects who have data only on Y0
3) The Initial Values file
The initial data file contains the initial values used for running the program. The initial values for beta/gamma parameters are varied when multiple chains are used. BOA program in R was used to calculate the univariate and multivariate Scale Reduction Factors (SRF).
Niko Kaciroti
300 N. Ingalls Building, 10th Floor
Center for Human Growth and Development
University of Michigan
Ann Arbor
MI 48109
USA
E-mail: nicola@umich.edu
|
