739109_en-US

liliconcom

calculate the total percentage of outliers for a process of several properties combined

I am trying to calculate total yield of a process. The process will decide several properties of a product. For any one of the product that comes out from this process, if any one of these properties falls outside the limits we target for,  then this product should be counted as a outlier. I found it convenient to calculate the outliers for each of the single property alone, but I did not find a way to calculate the percentage of the outliers that is defined by those which violate any one of the list of properties combined. Is there a way to do this for multi-property process? 

Basic Data Analysis and ModelingQuality and Process EngineeringRe: calculate the total percentage of outliers for a process of several properties combined

It would be helpful to see what your data table looks like, and how you are determining the different  properties for the products.

Re: calculate the total percentage of outliers for a process of several properties combined

Hi, Jim

Thanks for the reply. 

This is my table

Basically, I want JMP to compute the total percentage of entries that satisfy both the length and height criteria

 

	length (target 55, low limit 54.5, high limit 55.5)	height (target 14.5, low limit 13.5, high limit 15.5)
1	55.2118	12.8086
2	55.2468	14.8872
3	56.2425	14.8144
4	55.2735	14.8709
5	59.243	16.87

Re: calculate the total percentage of outliers for a process of several properties combined

Have you tried displaying the statistics associated with the Distribution platform?  If you enter the spec for each column, you can get numerous summary statistics.

 

https://www.jmp.com/support/help/en/17.2/?os=mac&source=application#page/jmp/process-capability.shtml%23

 

Commentary you can ignore....Your definition of "outlier" is the data point falls outside the "spec limits".  This is not a statistical definition of an outlier as the spec limits are derived independent of the true process variation.  Why not use the actual measured values instead of converting it to a categorical (e.g., nominal) value?  This would provide much more information.