JAGS

Working on simulation for a paper, I use JAGS fro cross-verification. 1. http://www-fis.iarc.fr/~martyn/software/jags/ get 1.03 package for windows it seems that 1.04 has been released but not available for windows yet. http://calvin.iarc.fr/wordpress/ 2. http://www-fis.iarc.fr/~martyn/software/jags/ for installation and example a short intro with an example www.stat.ufl.edu/~rsteorts/jags_present.pdf

logistic regression and classification/PS

Walker and Duncan (1967) longitudinal logistic classification model. 1. data collected in a 10-year period 2. number of variables is huge The observations are not independent to one another; model selection and variable reduction; Cornfield ( 1962) argued that "The use of a mathematical model which summarizes the observations in a small number of disposable parameters seems to offer the only present hope of obtaining quantitative answers to questions of interest". If the parameters are disposable, how would we know it is replicable and reliable? And how much the results (classification results) based on the disposable parameters are reliable? Is this, related to or set up the theoretical tone for, sample-based classification modeling (internal classification)?