tom@raal.UUCP (Tom De Pauw) (06/13/90)
Can anyone mail/suggest algorithms/packages (preferably in APL, but
FORTRAN or C will do) to analyze actual failures of equipment (Weibull
parameter estimation, etc). Also, can you recommend a good textbook on
the subject (& phone number for quick ordering).
Depending on response, I will summarize.
--
Tom De Pauw, R. Angus Alberta Ltd, Caterpillar (TM) dealer, Edmonton, AB
alberta!{aunro | adec23}!raal!tom
"Please no contact with present employer until after interview."tom@raal.UUCP (Tom De Pauw) (06/13/90)
Can anyone mail/suggest algorithms/packages (preferably in APL, but FORTRAN or C will do) to analyze actual failures of equipment (Weibull parameter estimation, etc). Also, can you recommend a good textbook on the subject (& phone number for quick ordering). Depending on response, I will summarize. -- Tom De Pauw, R. Angus Alberta Ltd, Caterpillar (TM) dealer, Edmonton, AB alberta!adec23!raal!tom "Please no contact with present employer until after interview."
chappell@yates.uchicago.edu (Chappell) (06/15/90)
I have some experience in fitting the Weibull distribution to data with various patterns of censoring and truncation. In many cases, it is quite easy to fit. There a few factors which must be considered before the modelling is performed: 1) Is there any censoring, or are all the failure times known exactly? 2) Is computational efficiency important (i.e., are you doing multiple analyses on huge data sets with a limited budget)? 3) What statistical packages, if any, do you have on your computer? For example, the SAS procedure LIFEREG fits Weibull models to censored data, as does GLIM (using the macro given in the second newsletter, and which is included in recent GLIM releases). 4) Do you know the index parameter, or do you want to estimate it? In the former case, the data can be transformed to be exponentially distributed, in which case estimation is trivial. A nice textbook which gives a very explicit example for fitting the Weibull using Newton's method (giving all the necessary derivatives and everything, enabling you to program it in a jiffy), is: Survival Analysis, by D.R. Cox & David Oakes, Chapman & Hall. Wayne Nelson, in a Wiley book published 5 or 10 years ago, gave a treatment which is more oriented towards engineers. Rick Chappell chappell@galton.uchicago.edu