Best Practices for Statistical Sampling of Damaged Goods

[CliffMichael]

I have a finished goods inventory of 700 automobile speaker/amplifier systems that were damaged in an overturned common carrier shipping trailer. I think it reasonable that most of the units can be recovered for salvage purposes. I propose to implement appropriate functional testing of a small sample quantity to statistically establish a probability of an acceptable level of functionality of the larger untested inventory.

Given that the units were factory tested and known to be functional but were exposed afterwards to moderate-to-severe G-force impacts; how does one select sensible criteria for a sample quantity to test in order to reach a given level of confidence for the larger untested inventory.

 

[melone]

It all depends on what the failures could be. I would suggest that you just test all 700, and then you will have very high confidence. Remember, the amount of damage could be proportional to the location in the trailor, i.e. what was on the top, now has the weight of all the other devices weighing down on it after the crash.

 

[DougMSOE]

Right from the statistical handbook.
a sample size for large populations might be approximated by the square root of the population, i.e. about 27.
Split the group into a random bunch of groups of e.g. 10 groups of 70.
What I have done is used a pair of dice, (sometimes as many as 6 die)
Using the dice pick a random sample to total 27.
test the ones that you picked and use the standard hypothesis testing.
the reason for the dice is to randomize the number that you skip and the groups that you pick from.