Gage Linearity and Bias Study Analysis - Minitab help needed

Hi,

Could someone help me analyze the results of the "Gage Linearity and Bias Study", please? :mg: Attached to this post are the measurements and results (graph). I don't know how to analyze these results!

I only know that the %Linearity variation corresponds to 8.8% of total variation (gage R & R study) and that the %Bias variation corresponds to 0.4% of total variation. Both p-value is 0.000, what does that mean?

Please help me with a complete analysis of the results. I´m newbie and I need your help !

Thanks in advance !

Best regards,
Licht

Licht said:

Hi,

Could someone help me analyze the results of the "Gage Linearity and Bias Study", please? :mg: Attached to this post are the measurements and results (graph). I don't know how to analyze these results!

I only know that the %Linearity variation corresponds to 8.8% of total variation (gage R & R study) and that the %Bias variation corresponds to 0.4% of total variation. Both p-value is 0.000, what does that mean?

Please help me with a complete analysis of the results. I´m newbie and I need your help !

Thanks in advance !

Best regards,
Licht

Click to expand...

Licht

The first thing I notice is tha the reference or master values are 4 significant figures but your measurements are only 3 significant figures.

Hi Licht,

There's always some danger in explaining what's going on just from the output without knowing much about how the data were collected, but I'll try. Some other things that might help:
If you right-click with your mouse, the context menu should give you access to StatGuide, Minitab's guidance that helps interpret what you're looking at.
Minitab also publishes an online explanation of the Gage Linearity and Bias Study. I don't have enough posts to include a link, so you'll have to copy and paste it, including www at the beginning:
minitab.com/en-US/support/answers/answer.aspx?id=482
I think that the one in StatGuide is more helpful. Fortunately, the online explanation gives alternative instructions for getting to StatGuide in case mine are unclear.

Here's my take:
Ideally, there would be no bias in your measurements, which would also guarantee that the amount of bias didn't depend on the size of what you were measuring. That is, ideally the black line that shows the best fit of the bias for your data would cover the horizontal blue dashed line.

Because your line slopes upwards, it is saying that, in your data set, your measurements are larger than the parts as you measure larger parts. Correspondingly, your measurements tend to be too small for smaller parts.

Of course, the reason the line is sloping up is primarily because, compared to the rest of your data, the measurements for the master part that is 2,458 are much larger than the real size. The line has to slope upwards to fit the average bias at that point. Because the bias for this master is at least 4 times larger than the bias for any other master, I'm inclined to think that you should first investigate the problem there.

Some good points have already been made, so I will try to add to them.

Start with the p-values. Low p-values (<= 0.05) mean that there is very low probability that you obtained these results by chance. Higher p-values would have meant that you could disregard bias and/or linearity. In your case you cannot disregard them statistically. However, you may be able to disregard them practically. However, that requires further evaluation.

First Steven's point. If your gauge resolution is inadequate it almost guarantees problems with any type of measurement system study. The question to answer is whether the statistical problems cause practical problems. They are not always the same.

Second, Steelejc's point. the results for 2.458 do seem anomalous and warrant further investigation. The preceding measurements do not show such an upward trend.

Lastly, the practical issue. Are you using this gauge for SPC or for inspection? While the Linearity may be 8.8% of the variation, it might also be 2% of the tolerance. If the gauge is used for inspection alone, there is no practical problem. Likewise the bias.

Hi everyone!

First, thanks for the answers. I think I understood your comments, but I need to share with you some more information to better understanding the procedures I have adopted.

I selected five pieces (8,7,14,15,1) to make the "Study of Linearity and Bias" with Minitab. These selected parts (electronic boards) were modified in laboratory to have values ​​(current values ​​[ampère]) within the range of specification, ie the part #8 was modified to have a value close to the low specification limit (1.97 A) and the part #1 was modified to have a value close to the high specification limit (2.45 A) and others (#7,#14,#15) within that range.
Subsequently, these modified parts were measured in the laboratory (with an oscilloscope) to obtain the master value (reference). So, I think the difference between parts are expected...right?

>> The specification limits are [1,95 A, 2,45 A]

Are my methods right?

regards.