I'm looking to perform a comparison study between two pieces of functional product test equipment. I want to determine how closely these testers provide me the same results from the same boards tested.

Some recommend using the GRR method -take 10 boards, 3 times, and vary the test station instead of the operator. Then use the reproducibility result to tell you the variation between the two testers.

Others have send to use the 2-sample t method - but use 1 board 30 times on each station. Process the data based on a desired 95% confidence interval to tell you how much of the UL-LL window is used by the variation between the testers.

Are these two methods really telling us the same things?

Is one better than the other?

Thanks,

slamers

I'm looking to perform a comparison study between two pieces of functional product test equipment. I want to determine how closely these testers provide me the same results from the same boards tested.You should always start out with stating the purpose of the study. Your statement above can be interpreted several ways. State your purpose clearly:

1) I want to determine whether the results of Test Unit #2 are statistically equivalent to the results of Test Unit #1.
2) I want to quantify the effect of Test Unit #1 and #2 on the measurement variation.

Some recommend using the GRR method -take 10 boards, 3 times, and vary the test station instead of the operator. Then use the reproducibility result to tell you the variation between the two testers.If your purpose is #2, this is the correct approach if the Test units are automated. If they are manual, you should use a more sophisticated DOE/ANOVA approach using multiple operators, test units, parts and repeated measurements.

Others have send to use the 2-sample t method - but use 1 board 30 times on each station. Process the data based on a desired 95% confidence interval to tell you how much of the UL-LL window is used by the variation between the testers.This approach would answer purpose #1. However, do not use an arbitrary quantity of 30. Decide in advance the difference in measure that is of practical importance to you. Use this to calculate the sample size needed to detect this difference, then use the hypothesis test or a similar approach such as a one-way ANOVA or one-way ANOM.

Are these two methods really telling us the same things?

Is one better than the other?

Thanks,

slamersNo these methods provide two different answers as discussed above. One method is not better than another, but one is more appropriate than the other depending on the purpose of the study.

:thanx:

Thanks. I think I understand enough to be a little more dangerous.

At first I wasn't sure exactly what I was seeking - I really thought of them as one in the same.