Shane: Just to add some practical advice on top of Rich's interpretation of the censoring process (which is correct), my recent experiences with analyzing below-detection limit chemical concentrations in water using left-censoring estimators indicates that historically people have not always indicated the below-detection limit data in a consistent fashion. So often people would include either a < with the detection limit value (e.g, <5), possibly also negative values (e.g., -5, negatives could be used as a flag for censoring if the real measurement scale is strictly positive as it is for chemical concentrations, but it has risky potential side effects), and possibly a separate column variable like your Censored TRUE/FALSE. I'm wondering if it is possible that your data format has a mix of these approaches. It seems like most recent statistical software for censoring usually wants a column variable for the measured response (Y), where observations of Y that are below (or above) detection-limit (censored) have the detection limit (censoring) value; and then a second column like your Censored column (either with TRUE/FALSE or 0/1 indicators).
Brian Brian S. Cade, PhD U. S. Geological Survey Fort Collins Science Center 2150 Centre Ave., Bldg. C Fort Collins, CO 80526-8818 email: ca...@usgs.gov <brian_c...@usgs.gov> tel: 970 226-9326 On Thu, Mar 14, 2013 at 9:20 AM, Rich Shepard <rshep...@appl-ecosys.com>wrote: > On Thu, 14 Mar 2013, Shane Carey wrote: > > Thanks for your reply. My data frame contains the value and a true/false >> to indicate whether they are censored or not. >> >> So I have something like: >> >> Data Censored >> -1.2 TRUE >> -5.5 TRUE >> 5 FALSE >> > > These negative values are actual values so do these have to be made >> absolute? >> > > Shane, > > If your data represent chemical concentrations of some sort then, yes, > the data must all be positive. You cannot have a chemical concentration > less > than zero. If the negative values are actual, then what is the reporting > level? And what do they measure that is really a censored value? > > The concept of censored data, whether right or left censored, is that > there is no way of knowing the actual numeric value. In right-censored > survival analyses (e.g., medical trials) the death of an individual is > recorded. But, when the study ends for whatever reason, there are still > subjects alive and there's no way of knowing how long after the end of the > study they die. Ergo, their age at death is unknown or censored. > > With left censored data such as chemical constituent concentrations in > air, water, or some other medium, there is a concentration below which the > instruments cannot distinguish it from noise. All we know is that the > constituent is present but its concentration is somewhere between zero and > the detection/reporting limit. Therefore, having a number that is below > this > detection/reporting limit is meaningless, and it cannot be negative. That's > why it is flagged as being censored. The cenmle() function assumes these > conditions to be true. > > Please keep this thread on the mail list so others can participate and > learn from the conversation. > > Rich > > ______________________________**________________ > R-help@r-project.org mailing list > https://stat.ethz.ch/mailman/**listinfo/r-help<https://stat.ethz.ch/mailman/listinfo/r-help> > PLEASE do read the posting guide http://www.R-project.org/** > posting-guide.html <http://www.R-project.org/posting-guide.html> > and provide commented, minimal, self-contained, reproducible code. > [[alternative HTML version deleted]] ______________________________________________ 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.
