GRR a case study.

New to statistics said:

For the repeatability the values must be on one spot of the line or across the line right?

What actions should be done to address the repeatability first, training perhaps ??

Click to expand...

All points lie on and across the 1:1 line. Part to part variation shows along the line and measurement error is perpendicular to the line, so PERFECT repeatability will be dots all on (across and along) the line.

Solving repeatability is not a simple thing - you must understand what is causing the repeatability problem. This is an engineering problem: it could be the part geometry, it cold be within part variation, it could be the measuring device, it could be technique…

New to statistics said:

If you have to make a presentation what you include (graphs, Minitab output, both) ?

Click to expand...

Never present anything you don’t understand and for which you have no viable conclusions. Never present a graph 0r table of statistical output that doesn’t clearly support your conclusions.

You are not ready to present anything

Hello New to Statistics,

It's always a challenge to convey these kinds of techniques on this forum, for this thread you laid out your intent for a GRR study but of course I realise that a lot of detail is omitted for whatever reason. My reply to you is based upon first principles and my experience stretching back many many years, it's your task to fill in the blanks and join the dots.

Ok, to the study, from the data I have:

10 parts
5 operators
2 replicates

First step- Construct or view the range chart

Any observations outside the upper control limit indicate a lack of consistency with the measurement process

Any statistic computed from this study is hypothetical and therefore unreliable until this is corrected

As this first step is so important, you will need to look at:

Broadly, the technique being used by the operators, observe during the study.
How the characteristic being observed is being measured- you mention s max, s min etc what are these, are they constructions, is there ambiguity involved in their derivations?
The screening technique being used in the study- for example is it possible that operators can remember their previous measurements, have the samples been randomised to minimise this effect etc etc
Anything that would contribute to the repeatability of the results.

FIX THIS PROCESS AND REPEAT THE STUDY - when the process is consistent it will be demonstrated by the range chart having no observations outside the upper control limit


Once you have completed step 1, which will be an achievement, certainly not a given, please report back and we can look at the other steps. Please bear in mind, again, I don't know all the context behind your study, but usually these studies are meant to convey what would be achieved in the real environment and situation for measuring these parts.

I hope this is helpful, good luck.

New to statistics said:

Please find attached

Click to expand...

what are the measurements? how is OD min and OD max determined? OD average? What is S? it is the standard deviation of something? for example does the operator measure the OD 10 times and then record the maximum value and the minimum value?

Without knowing what these things are and how they were obtained no statistical study will have meaning.

Bev D said:

what are the measurements? how is OD min and OD max determined? OD average? What is S? it is the standard deviation of something? for example does the operator measure the OD 10 times and then record the maximum value and the minimum value?

Without knowing what these things are and how they were obtained no statistical study will have meaning.

Click to expand...

Hello the parts are tubes seamless
OD stands for diameter
S stands for wall thickness
Gauge is a micrometer with digital step of 0.001mm
Measured each sample twice ( Max and min values are received from the gauge and the average value is calculated)

Miner said:

Notice the difference between Operator 1 (no repeatability) and operator 5 (decent repeatability). Also, note the poor reproducibility between them with Operator 1 having consistently higher results than Operator 5.

It's a good idea to review the data visually before diving into the statistics. In this case, you can see there is a problem that must be addressed before performing any statistical analysis.

View attachment 30103

Click to expand...

Miner how you add this 1:1 line in Minitab?

Create two new columns for the X,Y coordinates of two points that will form the desired line. Create a scatterplot, go into Edit mode, select Add Item > Calculated Line then enter the two columns (one being the X coordinates and the other being the Y coordinates).

Miner said:

The previous plots were square, but after adding the spec line, equal axes would have drastically shrunk the data points into a tiny area. Hard to have both. And those were not best fit lines. They are 1:1 lines.

Click to expand...

Hello Miner for the calculated lines.

At first (for the below data i received
X Y
10,019 10,019
10,027 10,027



For those
10,017 10,019
10,027 10,027



and last
10,017 10,019
10,026 10,027



in your graphs the diagonal line begins from the center of your charts
I believed (tell me if i am worng) that you must must put the same limit in Y & X axis to get the 45 angle right?
So the first diagramm is the proper one?
thank you

I find the below
I clicked the option Same... Y and X
and then insert the calculated line
Withe these values


i took the below result


i believe this is the proper approach what do you think???

@New to statistics You are correct. I did enter the same scale for both the X and Y axes. The calculated line appears to be correct.