Difference of two measurment methods
Six Sigma – iSixSigma › Forums › Old Forums › General › Difference of two measurment methods
- This topic has 8 replies, 3 voices, and was last updated 16 years, 2 months ago by
ww.
-
AuthorPosts
-
May 25, 2006 at 1:33 pm #43520
Can someone help me on this.
I have recently joined a company and they take inventory count by conducting a manual count first and then they scan the same items again. Any descrepencies are recorded and corrected on the final count sheet. I am trying to identify which method is more accurate.
What type of statistical test can I conduct to find out which method has proven to be more accurate?
Regards0May 25, 2006 at 2:08 pm #138095
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Joe,
You need to determine which is correct and then you test each distribution to a target – pairwise. The ideal difference is zero.
Testing them against each other will only tell you if they are thesame of different (not the correct statistical wording but it is how you end up treating the data).
Good luck0May 25, 2006 at 2:21 pm #138096Thank you Mike for the reply.
I am trying to justify droping one method of taking inventory but looking at the historical data and identify which has been more prone to errors. If I count the number of errors in the manual count and the number of errors in the scanned count, can I compare the two distribution of the two error counts and determine which has been more accurate in the past?
I hope I am making sense here. I apologize if not.
Regards,
Joe0May 25, 2006 at 2:43 pm #138099
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Joe,
If you compare them to each other all you will know is if they are the same or different. If you want to make the decision which is more accurate then you need to know what the real count was and compare the two methods (separately) to the actual (the issue becomes how do you establish which is the actual/target).
I have always looked at historical data out of the corner of my eye. Unless there was some really anal rentative person around you don’t know much about how or why it was collected.
I would set up a controlled test.
Just my opinion.0May 25, 2006 at 3:08 pm #138104Mike,
I have the real count and I know when they counted how many errors they had with the manual count and how many errors they had with the scanned count. (an error is when the count did not matcht the actual).
Knowing the number of errors on the manual count and the number of errors on the scanned count, do I just take the average of the errors accross many samples and assume that the lower average is more accurate?0May 25, 2006 at 3:17 pm #138110
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Joe,
You can but I would not do it. You have paired data run the test that way.
Just my opinion0May 25, 2006 at 3:26 pm #138114Mike,
Do I run a paired t Test? or do I establish a Hypothesis that the scanned test is more accurate and run a p Test?
Regards0May 25, 2006 at 3:33 pm #138117
Mike CarnellParticipant@Mike-CarnellInclude @Mike-Carnell in your post and this person will
be notified via email.Joe,
I would run it paired and the null hypothesis is that the difference is equal to zero.0June 1, 2006 at 2:56 pm #138502Just a thought; Human intervention and process are usually the major contributors where higher errors/nonconformances within any process and your inventory management system is a process. I would therefore tend to seek and rely on an automated system with a process plan that when followed, ensures success. If the plan is not followed or can not be followed then fix it so it can be followed and shoot the folks who won’t follow it (just kidding…well, your choice!). Good luck!
0 -
AuthorPosts
The forum ‘General’ is closed to new topics and replies.