[ | Next | Previous | Up ]

Re: Sampling plan for non-normal variable

From: Stan Hilliard
Date: 1/1/2005
Time: 9:22:14 PM

Comments

Hi Stan,

I also posted this in the April 2004 thread.

I was not very clear about "pick off two standard deviations from the normal chart."

To be more clear, I should have said:

Since a non-normal distribution is skewed, the estimate of the standard deviation (SD) that you use to calculate the upper acceptance limit will be different from the estimate of SD for the lower acceptance limit. For example, if the distribution is skewed to the high side, the upper one sided plan will use a larger standard than the lower side.

To estimate the standard deviation to use for the upper and the lower one-sided plans, use the fact that the SD for the upper side is the change in the variable from the 50 percentile to the 68.27 percentile. For the lower half of the distribution, SD is the change in the variable between the 31.73 percentile and the 50 percentile

Thus use the graph to calculate:

SD-upper = (X at 68.27 percentile) - (X at 50 percentile)

SD-lower = (X at 50 percentile) - (X at 31.73 percentile)

Then use these standard deviations to develop the respective upper and lower variables sampling plans.

If you develop the two plans to have the same sample sizes, you can express the decision rule limits as if they were one plan with an upper and lower limit.


Last changed: November 20, 2007