Two users ran across the same bug in icenReg v2.0.3: the quick summary is that if right censored data is coded as "NA", in some cases this confuses icenReg and the observations with NA's are dropped. THIS IS VERY BAD!!
At this time (10/1/2017), the bug has been patched, but the patched version is not on CRAN yet. I expect the patched version to be on CRAN within a week.
In the meantime...how do you get valid inference? The safest thing to do is to replace your NA's with Inf's. In otherwords, if your variable u is the upper end of your response variable in your data, simply apply
data$u[is.na(data$u)] = Inf
Alternatively, you don't want to edit your own code, you can install the patched version of icenReg from gitHub. To do this, call
library(devtools)
install_github('pistacliffcho/icenReg_devel/Code/icenReg')
I apologize for the error.
FOLLOW UP:
icenReg v2.0.4 is now on CRAN, in which this error was fixed.
At this time (10/1/2017), the bug has been patched, but the patched version is not on CRAN yet. I expect the patched version to be on CRAN within a week.
In the meantime...how do you get valid inference? The safest thing to do is to replace your NA's with Inf's. In otherwords, if your variable u is the upper end of your response variable in your data, simply apply
data$u[is.na(data$u)] = Inf
Alternatively, you don't want to edit your own code, you can install the patched version of icenReg from gitHub. To do this, call
library(devtools)
install_github('pistacliffcho/icenReg_devel/Code/icenReg')
I apologize for the error.
FOLLOW UP:
icenReg v2.0.4 is now on CRAN, in which this error was fixed.