Variable data treated as attribute

This topic has 4 replies, 4 voices, and was last updated 13 years, 2 months ago by annon.

Viewing 5 posts - 1 through 5 (of 5 total)

Author
Posts
December 10, 2007 at 3:47 pm #48887

claire
★★★★★★★★★★
Participant
@claire Include @claire in your post and this person will
be notified via email.
Hi, we are working on a process where we are getting variable data (i.e. actual numbers) but because of the measurement system, the number of discrete points is less than 5. In other words, when we measure diameter, we do not get a wide range of diameters, all diameters are between 0.121-0.124″ (this is as low as our measurement system will go). Therefore, should these data be treated as attribute, i.e. pass/fail, so long as they are in spec? We would normally do tolerance intervals on variable data, but given the small size of variability, we have decided not to do so. Is this correct?
0
December 10, 2007 at 4:41 pm #165950

Anand
★★★★★★★★★★
Participant
@Anand Include @Anand in your post and this person will
be notified via email.
Hi Claire, In my opinion just treating the data as discrete may not be the best thing to do as it may limit the information that data can throw to us. What I would suggest is to conduct Normality test for the data and if it is found normal, we should treat it as Continuous data only but if it is non normal we can go ahead with treating it as discrete. Discrete data would demand much bigger sample size than continuous data for the analysis.
0
December 10, 2007 at 5:45 pm #165953

annon
★★★★★★★★★★
Participant
@annon Include @annon in your post and this person will
be notified via email.
How many distinct categories are you showing and how did you determine this number?
0
December 10, 2007 at 6:33 pm #165955

Not Darth
★★★★★★★★★★
Participant
@Not-Darth Include @Not-Darth in your post and this person will
be notified via email.
Wow that is bad advice. Where did you learn such a thing?
0
December 10, 2007 at 10:19 pm #165962

annon
★★★★★★★★★★
Participant
@annon Include @annon in your post and this person will
be notified via email.
There is no reason why you can’t use an attribute form of your data, as long as you take the limitations into account. Pay attention to the number of distinct categories that your MS is capable of discerning, as this will determine what type of attribute data you can convert to…for example, 2 distinct categories will lend itself to a binary format, etc.
Good luck.
0
Author
Posts