Back to Basics Statistics - Standard Deviation small when performing an R&R

[Jstain1]

I have this dead horse. I'm told to kick it a few more times to see if I canget it to move... or so it seems.

I thought I'd try a different appraoch thi stime.

After performing an R&R I got the following data:

Lower st dev Upper
.0065249 .0152753 .236396

given 95% CI

I'm told this is no big deal, look at the StDev, it's small, no problem. Have a Nice Day.



I can't convince them that just because it's small doesn't mean it's OK

 

[Rob Nix]

One SD is small. 5.15 SD (99% or total estimated variation) is .0786678, which is 34% of total tolerance (.2304351), a bit high even by AIAG MSA reckoning.

 

[Darius]

I don't tink that Upper and Lower are Specs, It look more like confidence interval of something (may be stdev) because of the CI of 95%.

I don't tink it's R&R, Jstain1 can you clarify it?

 

[Jstain1]

they are the upper and lower limits in std dev's as set by MINITAB 14 given a 95% CI.

 

[Rob Nix]

they are the upper and lower limits in std dev's as set by MINITAB 14 given a 95% CI.

If that is the case, there is simply too little information here to determine whether the variation is acceptable or not. Please provide a little more background information.

NOTE: Darius, regarding your use of the word "tink", it should be "think".

 

[Darius]

Tanks Nix, I think it's my karma

Facilius est pro castitate, quam cum castitate mori.

Tertuliano

 

[Govind]

Jstain,
I just reviewed MiniTab Rel14 (our LAN version). I dont see this stddev CI in the R & R output (Both methods). Can you confirm which output was that? It would be useful for me to know the context of the analysis before providing a feedback.
Govind.

 

[Steve Prevette]

Two things that haven't been brought up yet:

1. What is the time series of the data? We can have a fine looking CI, fine looking standard deviation, but if the data are showing a trend (ie NOT STABLE), then it is meaningless.

2. To judge if the R&R spread is acceptable (assuming the data were statistically in control), you really need to know the application the gage is being used for. If the gage readings actually did vary as much as the R&R is telling you, what are the cost effects of that? Will parts that are outside of specification potentially fail to be detected because the gage is too sloppy? Even if in spec, will parts no longer fit together well enough? Or is this a case of "close enough" as in playing horseshoes (or hand grenades)?

 

[Jstain1]

ATE's

OK, let me see if I can clarify:

I have several ATE's which test electrical mteters. Just like the ones hanging outside your house. I got 5 of them from the same lot. They are enclosed and can not be adjusted. I used 3 ATE's, four trials and 5 parts. My intention was to see just how far apart they each were from each other. I tested at the same time each day. i randomized the testing. I was the operator, however I wanted to stay within the spirit of the MSA 93rd ed).

I'm using MINITAB 14, I used, Quality Tools, Gage R&R, crossed.

Gage R&R 

                              %Contribution
Source               VarComp   (of VarComp)
Total Gage R&R     0.0002072          61.10
  Repeatability      0.0001428          42.11
  Reproducibility    0.0000644          18.99
    trial                0.0000644          18.99
Part-To-Part        0.0001319          38.90
Total Variation      0.0003391         100.00


                                Study Var  %Study Var  %Tolerance
Source             StdDev (SD)      (6 * SD)       (%SV)    (SV/Toler)
Total Gage R&R       0.0143949    0.086369       78.17       28.79
  Repeatability        0.0119502    0.071701       64.89       23.90
  Reproducibility      0.0080253    0.048152       43.58       16.05
    trial                  0.0080253    0.048152       43.58       16.05
Part-To-Part          0.0114852    0.068911       62.37       22.97
Total Variation       0.0184153     0.110492      100.00       36.83

Number of Distinct Categories = 1



Now, the engineers here say, look the standard deviations are small. the process is good and the ATE's are doing a good job.

I say, no. the process isn't good and the ATE's are poor.


So who's correct?

 

[Sam]

Tanks Nix, I think it's my karma

Facilius est pro castitate, quam cum castitate mori.

Tertuliano

Translacion, por favor