I have a bore gage, one tip with 3 point contact for measuring a hole diameter, and one tip with 2 point contact for measuring over pin diameter on a spline. Both R&R's, and for that matter pretty much every bore gage R&R (these are fowler bore gages, the type that look like a gun) I've done the R&R is horrible. The contacts don't seem to allow for a very stable reading. Is there a better gage available in someone's opinion? We have a few dial bore gages, but these are somewhat impractical and are more difficult for the operators to use. Also, I have one more gage question. We make an output shaft. We receive the forging and perform some minimal machining, in essence we are providing a cleaned up forging to the customer. On the stem surface, occasionally there are nicks and gouges. Due to subsequent machining, we are allowed up to 0.03" on the nick. Currently we are using an indicator with a fixture, V shaped to measure the nick. This gage is also pretty bad, and is difficult to use because you must first zero it on the forging surface, and also cannot get to every area on the part (flange, etc.) Does anyone have any reccomendations on a good gage to use to check this depth? I thought of perhaps extending the fixture and making it longer, but the gage in and of itself it pretty bad and not repeatable.
Send to friendSign In to get started!
Comments
forrestbreyfogle 11/3/2004
INITIAL POSTING: I came across with a thread in another board where the user has used mean, standard deviation to analyze the data on Supplier Performance assessment values (Likert Scale 1-10). This triggered my thought and hence the following question. In your book "Implementing Six sigma" on Page 653, you had suggested to use Sixsigma tools like Normal probability plotting and ANOM to obtain insight on variability for data like Employee Opinion Survey. As indicated in the text, these are "Likert Scales". Hence the data is discrete. Would you please explain how statistical procedures relevant to Continuous distribution can be used for Likert Scale data? I incidentally came across with a webpage
http://www.icbl.hw.ac.uk/ltdi/cookbook/info_likert_scale/ that suggest median or mode to measure central tendency and range, intrerquartile range for variability. Also comparison of groups through Non Parametric methods. What are the applicable statistical procedures/analysis for Likert Scale Data?
RESPONSE: Your reference was made to the first edition of "Implementing Six Sigma". Within the additional 400 pages that are included in the second edition I discuss non-parametric tests too.
With a ruler you have the same distance between the 1 to 2 inch marking as the 4 to 5 inch marking. The problem with opinion scales is that you do not necessarily have the same "distance" between the units (i.e., 1 to 2 might not technically be the same as 4 to 5). Technically you should use non-parametric statistics to make comparisions.
However, non-parametic analyses do not typically give as much insight as a parametric look at the situation. Also, keep in mind with Six Sigma we are looking for insight, not necessarily the perfect model for situation.
Hence, I might "bend the rules some" to get a different view of a situation, remembering not to quote my finding to the third decimal place.
I am sure that some (perhaps many) will argue the point statistically, but mean analyses can be some what robust to these survey problems, especially when care is taken with how the survey is setup and analyzed. REMEMBER, I am only trying to offer another OPPORTUNITY for looking at the data; i.e., not a perfect solution.
Relative to your comment about probability plotting. This is more of a stretch relative to being statistical perfect; however, a probability plot could offer you some insight to variability in population response (keep in mind this is only a picture).
Please contact me directly if you would like to discuss further. Hopefully my response has not disturbed anybody too much.
Forrest Breyfogle
512-918-0280 x401
forrest@smartersolutions.com
www.smartersolutions.com
gramu 11/5/2004
Dear Mr.Breyfogle,
Thanks for the feedback.Here is an idea from Mr.Steve Prevette,ASQ CQE in the ASQ forum. He has referred to a method to convert Likert scales to an interval scale statistically.
http://www.hanford.gov/safety/vpp/survey.htm
Iam sharing the reply for the benefit of Inside Quality Forum users.Let me your thoughts about this approach.
Regards,
GR