Dear List, Here is an example of survival data in counting process format (detailed record of each day)
> data[data$Id == 11,] # extracted one person's record Id Event Fup Start Stop sex Drug1 601 11 0 6 0 1 0 0 602 11 0 6 1 2 0 0 603 11 0 6 2 3 0 0 604 11 0 6 3 4 0 0 605 11 0 6 4 5 0 1 606 11 1 6 5 6 0 1 which is compressed in the following format (unchanged records of drug exposure merged): > compressed.data[compressed.data$Id ==11,] # compressed same person's record Id Event Fup Start Stop sex Drug1 21 11 0 6 0 4 0 0 22 11 1 6 4 6 0 1 My question is: since the provided information is the same, should I expect numerically exactly same results from the following coxph outputs? If no, then which format is recommended? > data <- read.csv("http://stat.ubc.ca/~e.karim/dd.csv") > compressed.data <- read.csv("http://stat.ubc.ca/~e.karim/cd.csv") > head(data) > head(compressed.data) > coef(coxph(Surv(Start, Stop, Event) ~ sex + Drug1 + cluster(Id), robust = T, > data)) sex Drug1 0.8696213 3.1755854 > coef(coxph(Surv(Start, Stop, Event) ~ sex + Drug1 + cluster(Id), robust = T, > compressed.data)) sex Drug1 0.8674742 2.7147013 PS: discrete time analogue to Cox's (using cloglog link) also gives similar results corresponding to the dataset chosen. Any suggestions/references/direction to R-package will be highly appreciated. Thanks, Ehsan ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.