[R] glm-logistic on discrete-time methods with individual and aggregated data

From: Camarda, Carlo Giovanni <Camarda_at_demogr.mpg.de>
Date: Thu 02 Feb 2006 - 05:31:21 EST

Dear R-Users,
without going into details I tried to prepare a simple example to show you where I would need help.
In particular I prepare two examples-template for a study I'm conduction on discrete-time methods for survival analysis. Each of this example has two datasets which are basically equal, with the exception that in the former one has individual data and in the latter one aggregated data.
The difference between the two examples is on a single subject: I substituted to the first example a censored case with a subject who died at the first time-unit.
Afterward I fitted a logistic model (Fahrmeir and Tutz, 2001) in the glm context, but whereas there is not difference between individual and aggregated dataset in the first example, I noted some discrepancies in the second example. I might guess that something with weights is going on, but I did not manage to clearly understand. Hope that the following example will be more clear than my explanations, Thanks in advance,
Carlo Giovanni Camarda

rm(list = ls())
# working one

timesIND <- c(rep(1:4, 3), 1, rep(1:2,2), rep(1:3 , 2), rep(1:4, 2))
statusIND <- c(rep(0 ,12), 1, rep(0:1,2), rep(c(0,0,1), 2), rep(c(0,0,0,1),2))
datiIND <- as.data.frame(cbind(timesIND, statusIND)) datiIND$timesIND <- as.factor(datiIND$timesIND)

timesAGG  <- c(  1:4,    1,   1:2,      1:3,       1:4)
statusAGG <- c(rep(0,4), 1,   0:1,    c(0,0,1), c(0,0,0,1))
weightAGG <- c(rep(3,4), 1, rep(2,2), rep(2,3), rep(2,4))   
datiAGG <- as.data.frame(cbind(timesAGG, statusAGG, weightAGG)) datiAGG$timesAGG <- as.factor(datiAGG$timesAGG)

coef(glm(statusIND ~ timesIND, family=binomial, data=datiIND)) coef(glm(statusAGG ~ timesAGG, family=binomial, data=datiAGG, weights=weightAGG))

# not working one

timesINDa <- c(rep(1:4, 4), rep(1:2,2), rep(1:3 , 2), rep(1:4, 2))
statusINDa <- c(rep(0 ,16), rep(0:1,2), rep(c(0,0,1), 2), rep(c(0,0,0,1),2))
datiINDa <- as.data.frame(cbind(timesINDa, statusINDa)) datiINDa$timesINDa <- as.factor(datiINDa$timesINDa)

timesAGGa  <- c( 1:4,        1:2,     1:3,      1:4)
statusAGGa <- c(rep(0,4),    0:1,   c(0,0,1), c(0,0,0,1))
weightAGGa <- c(rep(4,4), rep(2,2), rep(2,3),  rep(2,4))   
datiAGGa <- as.data.frame(cbind(timesAGGa, statusAGGa, weightAGGa)) datiAGGa$timesAGGa <- as.factor(datiAGGa$timesAGGa)

coef(glm(statusINDa ~ timesINDa, family=binomial, data=datiINDa)) coef(glm(statusAGGa ~ timesAGGa, family=binomial, data=datiAGGa, weights=weightAGGa))

This mail has been sent through the MPI for Demographic Rese...{{dropped}}

R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html Received on Thu Feb 02 05:49:34 2006

This archive was generated by hypermail 2.1.8 : Fri 03 Mar 2006 - 03:42:16 EST