Gage R&R: Gage Sigma > Total Sigma.......??

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by PMH (via 194.106.141.243) on November 13, 2001 at 6:07 AM (ET)

Hi all,

Having performed a Gage R&R study I find that the variation due to the Gage (as derived from the Gage R&R) is greater than the total variation (Std Deviation) of the process. How does one intrepret this?

Is it possible for the Gage capability to be larger than the the total variation from the same data set?

All help would be greatly appreciated.
Paul.

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by Darius (via 200.23.217.10) on November 13, 2001 at 12:57 PM (ET)

The problem must be wrong "rational sampling", if it is wrong it may appear the within sample variation greater that the total variation.

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by PMH (via 194.106.141.243) on November 13, 2001 at 1:56 PM (ET)

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by PMH (via 194.106.141.243) on November 13, 2001 at 1:57 PM (ET)

The study was a standard Gage R&R (10 random samples x 3 meaurements each x 3 operators). How does rational sampling apply to this?

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by Darius (via 200.23.217.10) on November 14, 2001 at 11:20 AM (ET)

The principles for subgrouping are:
1- Never knowingly subgroup unlike thing together.
2 - Minimize the variation within each subgroup
3 - Maximize the opportunity for variation between the subgroups
4 - Average across noise, not across signals
5 - Treat the chart in accordance with the use of the data
6 - Establish sampling procedures.

The most common error in subgrouping is principle #3, so one has to make the cuestion "Wich sources of variation ocurr between the subgroups?"

So: the tree samples must be in the most different values for the variable and each sample must be tested by the operators (the same one).

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by Darius (via 200.23.217.10) on November 14, 2001 at 12:08 PM (ET)

Sorry, I try to say

So: the samples (10) must include the possible range for the variable (ie. if it is a pH and the process can behave between 6.5 and 10, being the specifications between 7.5 and 8.5, the samples must include between 6.5 and 10) and each sample must be tested by the operators (the same one).

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by Phil (via 66.32.157.178) on November 15, 2001 at 11:30 PM (ET)

The standard deviation should include the variation from the process PLUS the measurement variation (all variation). Based on that, the gage variation should not be more than the standard deviation...unless you made a mistake.

I may not understand your situation, but that is my guess.

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by PMH (via 194.106.141.243) on November 16, 2001 at 4:29 AM (ET)

Thanks Darius, Phil.

What I am trying to analyse with the Gage study is a test set-up for testing Radio-Frequency antennas. RF work is very prone to fluctuations.

I am inclined to go with the rational subgroup theory as it was observed during the study that the Range in measurements within one subgroup (i.e. one antenna measured three times by the one operator) was often larger than the group-to-group variation (i.e. the range between the average ranges of each operator).

I am going to repeat the Gage study, but this time take each data point to be an average of the fluctuating RF signal rather than a snapshot. This should reduce the within group range and hopefully solve the problem.

Does anyone feel that this is a bad move?

Phil, the total standard deviation and the Gage deviation were calculated from the same data set. The total the std deviation of all the data points, the Gage deviation from the Sqrt(Repeatibility^2 + Reproducibility^2). The maths has been checked rigourously.

Re: Gage R&R: Gage Sigma > Total Sigma.......??
Message posted by Darius (via 200.23.217.10) on November 16, 2001 at 1:38 PM (ET)

It may work, especially if the signal has too much variation at a given time.

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.