P
pmccain47
Hello All,
I have a statistics question that I was hoping someone might have some insight into. For one of our design criteria (pull force) we have a minimum pull force of 1.12 lbf. In design verification, we typically perform a 95% confidence/99% reliability analysis to demonstrate that the lower bound is above our minimum specification.
Usually we have no issues with that, but in a data set we are analyzing, we have a relatively consistent mean of ~2.5 lbf, but two samples that pulled significantly above that (3.3 lbf and 3.5 lbf). There is no max spec, the higher the pull force the better. But this variation to the high side increases the standard deviation significantly and causes our lower bound to fall at 1.05, which is below the min spec.
If the samples are excluded, the LTL is ~2.0 which would meet our acceptance criteria. There is no immediately assignable special cause for the variation, the units look identical to the ones that were closer to the mean. Are there any other strategies to account for variation above the mean of a one sided spec?
Thanks for any help or thoughts!
Please - SCAN the attachment for viruses or whatever.
.
I have a statistics question that I was hoping someone might have some insight into. For one of our design criteria (pull force) we have a minimum pull force of 1.12 lbf. In design verification, we typically perform a 95% confidence/99% reliability analysis to demonstrate that the lower bound is above our minimum specification.
Usually we have no issues with that, but in a data set we are analyzing, we have a relatively consistent mean of ~2.5 lbf, but two samples that pulled significantly above that (3.3 lbf and 3.5 lbf). There is no max spec, the higher the pull force the better. But this variation to the high side increases the standard deviation significantly and causes our lower bound to fall at 1.05, which is below the min spec.
If the samples are excluded, the LTL is ~2.0 which would meet our acceptance criteria. There is no immediately assignable special cause for the variation, the units look identical to the ones that were closer to the mean. Are there any other strategies to account for variation above the mean of a one sided spec?
Thanks for any help or thoughts!
Please - SCAN the attachment for viruses or whatever.
.