New user wants to predict %25 failed at certain %23 of installations

[Chad Mittelstadt]

[Originally Posted: 10/26/00--Transferred by ReliaSoft Moderator]

I am a new user with a brief background in distributions. I have data on the number of installations it takes to fail a sample. I would like to make a Weibull plot to extrapolate %'s from each side of the failure data. I am trying to follow a classic example from Reliability Toolkit: Commercial Practces Edition (RAC), but I am not sure if this is a valid application? If it is, what is the best technique to fit a line to the data points on the Weibull paper?

Thank you in advance,

[ReliaSoft Tech Support]

[Originally Posted: 10/30/00--Transferred by ReliaSoft Moderator]

I believe that the example provided in the RAC Toolkit (p. 170) would be applicable provided that the "number of installations" you mention is the time scale for your failure times, and that you have what is known as complete data. By complete data, we mean that all of the data points you have are those for failures where the exact failure time is known. If part of your data contains unfailed operating times (suspended data), or uncertainty of the actual failure time (interval data), the method outlined in the Toolkit is not valid, and you will have to use other means (such as ReliaSoft's Weibull++) to analyze your data.