Yes, that makes sense, and I hadn't thought of it -- I was thinking in terms of a nonparametric estimate of the hazard function. Spencer Graves makes a similar point. Andy Liaw was kind enough to point out to me that the muhaz function is in the muhaz package. As it turns out, muhaz provides smooth nonparametric estimates of the hazard function using kernel methods.
Thanks to all, John
At 03:06 AM 11/5/2003 +0000, Ted Harding wrote:
On 05-Nov-03 John Fox wrote: > Dear Monica, > > I'm not sure what the muhaz function is (it's not in the survival > package), but regardless, unless I'm seriously mistaken, there's no > information to estimate the hazard function if you haven't observed > any events. > > I hope that this helps, > John
Well, there is _some_ information, to the extent that such data rule out high levels of hazard ...
I recall seeing a paper by I.J. Good many years ago (can't locate the reference now) in which he made a Bayesian inference of the probability of nuclear war (none having occurred).
Basically he assumed a homogeneous Poisson process of nuclear war, with improper prior (? 1/mu ) for the mean, and got a posterior distribution for it. Consequently a probability of NW within the next (say) 20 years could be evaluated (though I seem to remember th\t a certain amount of footwork was involved).
In the present case, without going so far as to be Bayesian, assuming a constant hazard lambda would lead to an upper confidence limit for lambda given that there had been no events within the observed intervals (e.g. as the largest value of lambda such that the probability of no events was not less than 0.05). You don't need survival-data techniques for this ...
However, I certainly agree with the above to the extent that there is no information which would support an estimate of a non-constant hazard function.
Best wishes, Ted.
-------------------------------------------------------------------- E-Mail: (Ted Harding) <[EMAIL PROTECTED]> Fax-to-email: +44 (0)870 167 1972 Date: 05-Nov-03 Time: 03:06:44 ------------------------------ XFMail ------------------------------
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
----------------------------------------------------- John Fox Department of Sociology McMaster University Hamilton, Ontario, Canada L8S 4M4 email: [EMAIL PROTECTED] phone: 905-525-9140x23604 web: www.socsci.mcmaster.ca/jfox
______________________________________________ [EMAIL PROTECTED] mailing list https://www.stat.math.ethz.ch/mailman/listinfo/r-help
