PDA

View Full Version : Sample Size


B Karthikeyan
December 8th, 2000, 05:49 PM
[Originally Posted: 10/19/00--Transferred by ReliaSoft Moderator]

A product is failing at the rate of 195/1000 in the warranty Period ( Warranty Period - 12 months or 12,ooo Kms). Most of the products are failing between 5000-7000Kms. An improvement has been done to reduce the failure. Now how many samples to be tested & for how many kms inorder to verify the effectiveness of improvement.There is no accelerated test available. Pl send your replies to "bkarthikeyan@eth.net" or "bkarthikeyan@eicher.co.in"

Dr. Dave Olwell
December 8th, 2000, 05:50 PM
[Originally Posted: 10/21/00--Transferred by ReliaSoft Moderator]

Your question is a classic one, and has no easy answer unless you have an idea ahead of time about how big the improvement is going to be. Here is why.

If you have a very small improvement, it is going to take a large sample to demonstrate it convincingly. If you have a tremendous improvement, it will not take many samples at all to demonstrate it.

If you have an idea how large the improvement is, please post it along with the distribution you are using to model your failure times and we can try to provide some specific guidance.

Dr. Dave