Dr. Dave Stimley

Try some basic observational studies of the various gurus to see which produces the best results.  ANOM – Analysis of Means is an excellent method for accomplishing this.  Next, incorporate these best practices into a proceedure. Bring this procedure into a state of statistical control.  Then improve it if necessary or move on to the next pr…[Read more]

The best way of handling non-normal data when doing a capability study is to use a Pearson transformation which looks at the percentiles in the tails based on skewness and kurtosis and then gives the equivalent Cp and Cpk values.  (Reference Quality Progress – Sept. 1989).  This is also done automatically if using statistica software.  In any case…[Read more]

Use these formulae:  RTY= e^-(Sum of DPU) or -ln(RTY) = Sum of DPUs whichever is more convienient. 
Then DPMO = (Sum of DPUs*1000000)/sum of OE’s
Then Convert your DPMO’s to Sigma.
To get to Cpk: Once you know DPMO,  Take the DPMO value and divide it by 2 then find the Z value from a Z table.  For Example, 2700DPMO/2 = 1350DPMO or as a fra…[Read more]

Yes – I used to work for Motorola, used Statistica there as well.

Here is how it is done:
You will need the values of CPU and CPL to start.
Then:
1. Compute Zu = 3xCpu and Zl = 3xCpl
2. Find the fraction defective for each of the Z values from a Z table and add them together.
3. Multiply step 2 by 1,000,000 to get equivalent DPMO
4. From the DPMO you may find the sigma from a conversion table or…[Read more]