Sample size for life data determination

[Leiden Time]

Juran's Quality Control Handbook Section 12-12 tells that a "well chosen sample" will provide adequate "field intelligence". (More info can be found starting page 25-18 re Reliability Sampling)

Q1: Oftentimes it is time consuming and not economical to collect all times-to-failure/suspension data, so how do I determine the sample size (not for product testing purpose)from the bunch units actually put in service for life analysis? - say from a population of 100, 500,1000? What would be the considerations?

Q2: Juran's book relates TR (TR-3/-4/-6; 1961, 1962, 1963 respectively) to variable sampling (MIL Std 414), is not the latter designed mainly for acceptance inspection? Are TR-3-4 -6 also applicable to units put in service? Appreciate your comments.

Q3: I have hard time to locate these TR's at Government Printing Office web site. Any info be appreciated.

[ReliaSoft Tech Support]

The rule of thumb I like to follow is to collect as much reliability data as is practically or financially possible. As far as I am aware, there is no "magic number" for reliability sample size, despite some methodologies which claim to be able to provide fantastic results with tiny sample sizes. In general, the mare data that is collected, the more precision you will have in your results, and the tighter your confidence bounds will be. While this "more is better" is a rather vague answer to your specific question, you could see Montgomery's _Introduction to Statistical Quality Control_ for more information about sample sizes and sampling plans.