Revision Date: 2005-09-06
Problem:
At the 99.5% Confidence level, we say that the data fits the (bell) curve.
But, at the 99% level, we say that it does not. Are we actually saying the wrong (or confusing) words?
Solution:
The confusion is semantical. In the 99.5% confidence case, we would be saying that the data does not disagree with (negate) the normal assumption and therefore can be treated as if it is normal. In the 99% confidence case, the data does disagree with the normal assumption and therefore should not be treated as normal. In general this means that using the normal to figure theoretical percentages out-of-spec is incorrect. If there is sufficient data, the out-of-spec can be estimated using the actual percent out. In using the actual in order to predict out-of-spec, it would be nice if the process were in control during the period when the data is collected.
Gordon Constable, Ph.D. | |
