I am trying to estimate the kernel density for number of days a child is sick. Around 73% of children report not being sick, i.e. zero. How do I estimate a kernel density for this censored variable using Stata?

  • $\begingroup$ is there a particular reason that you're interested in a nonparametric model? since this is a discrete problem, it seems like there are plenty of good parametric choices. $\endgroup$ – Daniel Johnson Sep 27 '13 at 18:44
  • 5
    $\begingroup$ Calling this "censoring" implies you believe that some proportion of those reporting not being sick actually were sick but that some mechanism prevented the true positive values from appearing and replaced them with zeros. Usually it also means you know which observations were censored in this manner. Is this what's really going on, or are you using "censored" merely to mean there is a natural lower bound of zero to the values and you would like the KDE to respect it? If that's the case, your question has been answered at stats.stackexchange.com/questions/65866. $\endgroup$ – whuber Sep 27 '13 at 18:50