# Distinct Categories
Six Sigma – iSixSigma › Forums › Old Forums › General › # Distinct Categories
 This topic has 8 replies, 9 voices, and was last updated 16 years, 2 months ago by faceman.

AuthorPosts

January 17, 2005 at 2:36 pm #38097
I just ran a GR&R in minitab and my distinct categories = 1.now i know that is bad but im not certain why it is bad or if there is a common cause of the problem. We are still learning and trying to trouble shoot this MSA.Thanks for the help.
0January 17, 2005 at 3:22 pm #113593This means that your measurment system is worse than a pass/fail system (n of dist. cat. = 2). There is no resolution among measurements.
0January 17, 2005 at 3:40 pm #113594Good answer before. Plus, Minitab has a very thorough description of distinct categories with the math to explain the overlapping confidence interval analogy. You might also verify there is enough discrimination in your measurement system by reviewing a control chart using the range of variation within each person’s measurements. Look for 4 or more different values on the Y scale of the range chart in addition to using distinct categories.
0January 17, 2005 at 3:48 pm #113597newb,do your parts cover a great area of production range? If there is “no” spread in your data, your measurement system has a hard time.
Dieter0January 18, 2005 at 4:14 am #113636This is the way it was explained to me….I am sure some will think it is too simple.
Basically if you take two die (dice whatever) and say anything you roll from 16 is a one and anything from 712 is a two. Roll it and count how many “ones” and “twos” you have. Guess what you have two distinct categories. Now roll the same dice again and everything from 14 is a “one” from 57 is a “two” and 812 is a “three”. Now you have 3 distinct categories. You can do this grouping for everything up to 12 distinct categories. The more categories the information you have about the distribution. Now if I told you that if you rolled a 3 or an 8 it was a defect, how many distinct categories would you need in order to have a good measurement?
…two wouldn’t tell you anything, three is little better but you would still throw alot of good parts (rolls) away. You get the idea….
J0January 18, 2005 at 7:40 am #113641Nos. of Distinct Categories is the measure of resolution power of a Measurement system. Higher the nos., better is the resolution power i.e. measurement system is capable of differentiating two data points very close to each other. Finally higher is the least count of the measuring instrument, more will be nos. of distinct categories.
regards
dinesh
0February 1, 2005 at 1:12 pm #114290I have just completed a Gage RnR in Minitab and got “No. of distinct categories = 0”. and Gage R&R = 96.81%
Oh No!
We had three operators measure the same response twice. We carried this out 10 times.
Later, we sat around the table scratching our heads and decided to introduce lasers and all sorts to measure a critical dimension. “Yes lets go and spend money and get some fancy gear in!”, then our Gage RnR will come down, and distinct categories go up.
Then someone suggested we review our process capability before we rush off and perfect a measurement system that at worst is telling us we pass our customers criteria anyway !
Just a thought ?
0February 1, 2005 at 2:02 pm #114293All of the answers you got are good. here is a little mathematical explanation.
You calculate ndc as follows: (5.15*SigmaProcess)/(2.575*1.414*Sigmameas)The numerator is fairly easy to understand. The denominator is basically the discriminatory power of the gage. Lets say you make 2 measurements X1 and X2. Now the absolute difference between X1 and X2 must be greater than the denominator for the gage to distinguish the two parts. But the denominator as such is useless (ofcourse assuming the resulting value to be > 1) since we dont know how bad the gage is until we relate it to the process variation. Hence the ndc concept.0February 1, 2005 at 2:26 pm #114294
facemanParticipant@faceman Include @faceman in your post and this person will
be notified via email.If you have 0 ndc and an 97% R & R, how are you going to know if you have improved your process? How do know that this problem really exists if you are a measurment system with an R & R of 97%? Your measurement error will be confounded in the ‘process data’. The ‘observed’ process variance will really be process variance + measurement system variance.
Regards,
faceman0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.