From: Stan Hilliard
Date: 07 Oct 1999
Time: 13:35:40
Greetings John,
If you were to combine the data from the two temperatures to make one decision, you would have a bimodal distribution. This is assuming that temperature has, or might have, an effect on the performance variables. Z1.9 is based on the assumption that the variability amoung items follows a normal distribution - not bimodal. I would make the decision individualy for each temperature.
The other issue what sample size to use at each temperature, x or x/2. The complete accept/reject performance of each plan is defined by its oc curve. So plot and list the oc-curve for both plans, and compare them. See an example at:
www.samplingplans.com/modern3.htm#COMPARE
If you want to use software to do this, TP414 makes it easy. You just enter any n and k (plus the standard deviation for sigma-known plans). The software generates the oc curve plot and table.
Since the test is destructive, you can achieve a savings in sample size without modifying the chosen oc curve by applying a sequential sampling plan. Again, TP414 will generate the sequential decision rule. See examples of TP414 at:
www.samplingplans.com/programtp414.htm
One last point. If you have a situation where the mean is the most important statistic -- rather than fraction of items nonconforming to specification -- you could specify the acceptable and rejectable means directly with TP414.
Sincerely, Stan Hilliard