[R] Servreg $loglik

Charles Annis, P.E. charles.annis at statisticalengineering.com
Tue Jul 20 17:20:15 CEST 2010


Dear R-experts:

I am using survreg() to estimate the parameters of a Weibull density having
right-censored observations.  Some observations are weighted.  To do that I
regress the weighed observations against a column of ones.

When I enter the data as 37 weighted observations, the parameter estimates
are exactly the same as when I enter the data as the corresponding 70
unweighted observations.  This is to be expected, of course.

I don't understand, however, why the reported loglikelihood is 
> parameter.estimates$loglik
[1] -120.4699 -120.4699
for the 37 weighted observations, but
> parameter.estimates$loglik
[1] -135.1527 -135.1527
for the 70 unweighted observations.

(For the record, my computations of the loglikelihood, using the dweibull()
function for the observations and the pweibull() function for the censored
observations, is -135.1527 for both 37 weighted and 70 unweighted.)

I am using the data from Meeker and Escobar, _Statistical Methods for
Reliability Data_, Wiley (1998), Table C.1, shown below:

Hours	Status	Num.Parts
  450	Failure	1
  460	R-Censored	1
 1150	Failure	2
 1560	R-Censored	1
 1600	Failure	1
 1660	R-Censored	1
 1850	R-Censored	5
 2030	R-Censored	3
 2070	Failure	2
 2080	Failure	1
 2200	R-Censored	1
 3000	R-Censored	4
 3100	Failure	1
 3200	R-Censored	1
 3450	Failure	1
 3750	R-Censored	2
 4150	R-Censored	4
 4300	R-Censored	4
 4600	Failure	1
 4850	R-Censored	4
 5000	R-Censored	3
 6100	R-Censored	3
 6100	Failure	1
 6300	R-Censored	1
 6450	R-Censored	2
 6700	R-Censored	1
 7450	R-Censored	1
 7800	R-Censored	2
 8100	R-Censored	2
 8200	R-Censored	1
 8500	R-Censored	3
 8750	R-Censored	2
 8750	Failure	1
 9400	R-Censored	1
 9900	R-Censored	1
10100	R-Censored	3
11500	R-Censored	1

I am running R version 2.11.1 (2010-05-31) on a HP Windows 7 box with 8 gig
RAM.


Thank you for your help.

Charles Annis, P.E.

Charles.Annis at StatisticalEngineering.com
561-352-9699
http://www.StatisticalEngineering.com



More information about the R-help mailing list