Resolved: The problem was that the is_censored vector was not being passed in to jags with the other data.
Update: Strangely using the analogous model and OpenBugs, I recover what I expect. For right-censoring, the equivalent BUGS model replaces the dInterval call with the I(t_censored[i],) : model { # is_censored is 1 for censored values and 0 for non-censored values. # is_censored is modeled: isCensored ~ dinterval(y, censorLimit) # When is_censored is 1 and y is missing, then all JAGS knows is that y is # somewhere above censorLimit, so JAGS imputes a random value for y from # the likelihood function...
I simualte data that has a control and treatment group with an exponential distribution with median survival times of 30 and 40 respectively. These times imply rates of -log(0.5)/30 = 0.0231 and 0.01733. I generate time to event for both groups using rexp and generate censoring times in an analogous fashion. I am unable to recover the medians from the bugs model. See code below (and attached). set.seed(31) n_per_grp <- 250 grp <- c(rep(0, n_per_grp), rep(1, n_per_grp)) median_surv_time <- c(rep(30,...
I simualte data that has a control and treatment group with an exponential distribution with median survival times of 30 and 40 respectively. These times imply rates of -log(0.5)/30 = 0.0231 and 0.01733. I generate time to event for both groups using rexp and generate censoring times in an analogous fashion. I am unable to recover the medians from the bugs model. See code below (and attached). set.seed(31) n_per_grp <- 250 grp <- c(rep(0, n_per_grp), rep(1, n_per_grp)) median_surv_time <- c(rep(30,...
I simualte data that has a control and treatment group with an exponential distribution with median survival times of 30 and 40 respectively. These times imply rates of -log(0.5)/30 = 0.0231 and 0.01733. I generate time to event for both groups using rexp and generate censoring times in an analogous fashion. I am unable to recover the medians from the bugs model. See code below (and attached). set.seed(31) n_per_grp <- 250 grp <- c(rep(0, n_per_grp), rep(1, n_per_grp)) median_surv_time <- c(rep(30,...
I simualte data that has a control and treatment group with an exponential distribution with median survival times of 30 and 40 respectively. These times imply rates of -log(0.5)/30 = 0.0231 and 0.01733. I generate time to event for both groups using rexp and generate censoring times in an analogous fashion. I am unable to recover the medians from the bugs model. See code below (and attached). set.seed(31) n_per_grp <- 250 grp <- c(rep(0, n_per_grp), rep(1, n_per_grp)) median_surv_time <- c(rep(30,...