夏宁DOE 之 Paired Comparison

Identify eight Very good parts and eight Very bad parts which are having the problem that is under analysis.
. List as many Product characteristiscs/parameters which may explain the difference between the Good and Bad parts (This list is based on the engineering judgement of the person/team). List it in the descending order or importance to the problem.
. Measure all the Good parts and bad parts for the parameters identified above. There will then be totally 16 values.
. Arrange the 16 values in the rank order (start from the smallest to the largest or the largest to the smallest) irrespective of whether they are good or bad.
. Write against each value whether that particular value corresponds to a bad or Good part. If it belongs to the good part write ‘G’ within bracket after the value. If it belongs to the bad part, write ‘B’ within bracket after the value.
. From the top, find out when for the first time, the Good changes to Bad (or) the Bad changes to Good.
. Draw a line at the change point (For eg., if the change occurs after the 5th data, draw a line after the fifth data).
. From the bottom, find out when for the first time, the Good changes to Bad (or) the Bad changes to Good.
. Draw a line at the change point (For eg., if the change occurs after the 15th value from the bottom, draw a line between 15 & 16).
. Find out the top count (Count the number of values till the line is drawn).
. Find out the Bottom count (Count the number of values till the line is drawn from the bottom).
. Add both these counts to find the Total count.
. If the Total count <=5, then that quality parameter or product characteristic is not the reason for the problem.
. If the Total count is = 6, it can be concluded that at 90% confidence level, it is this quality parameter or product characteristic that is leading to the problem.

Confidence level for other counts

Number of Total count Confidence level
6 90%
7 95%
10 99%
13 99.90%