PDA

View Full Version : RRX and MLE confusion

keithtbe
January 14th, 2003, 12:26 AM
It is so confused when the beta is ranging between 0.9 to 1.1 while using RRX and MLE. Which is the most correct? If i have 300 data points with suspension data, it is always advice to use MLE. Is that correct or i should have to look at other criteria?
Pantelis
January 16th, 2003, 04:25 PM
With a large sample size (and you have one with 300 points) MLE should be preferred.
keithtbe
January 23rd, 2003, 06:25 PM
May i know what is the reason?
keithtbe
January 28th, 2003, 01:53 AM
May i know the reason?>
RS Support
January 30th, 2003, 04:52 PM
It’s not a simple answer - with people having strong feelings either way – however here it goes in a nutshell…
MLE estimates for the Weibull shape parameter tend to be biased for small sample sizes, thus some people have issues with this… on the other hand MLE estimates have much better properties than RRX estimates as well as handle censored data better …

Now with a larger sample size that you mention the biasness issues is not a factor… thus MLE

These are some related articles you may want to look at:
http://weibull.com/hotwire/issue1/hottopics1.htm
http://www.reliasoft.com/newsletter/4q2001/accelerometer.htm
http://www.weibull.com/hotwire/issue9/relbasics9.htm
http://www.reliasoft.com/newsletter/1q2001/simumatic.htm