July 28, 201312 yr If Data is not normal then these are the methods used to calculate process capability 1. Convert data to normal by BOX COX or JOHNSON transformation OR 2. If the best fit distribution (Stat>Quality Tools>Individual Distribution Identification) consider highest P-value distribution as best fit. If data neither fit in any distribution and nor transform to normal then 3. Consider continuous data as attribute data 3.1 Inspect each characteristic in your sample and identify as defective or non defective. Then calculate %defective. Using sigma level calculator calculate Z value and divide by 3 to calculate Cpk. 3.2 If No defects are observed in the sample to categorize as defective and non defective. Then calculate estimated proportion defective (Probability of defective)p(d)=1/(n+2) where n is number of characteristics inspected (TOFE), get Z by using one sided Z table and then divide by 3 to get Cpk.
August 29, 201312 yr Hi Sagar. Thanks for the above explanation. It is certainly useful. Can you throw some light on Johnson Transformation? Just the basic explanation would suffice.
Create an account or sign in to comment