When to use Linearity and Bias - appropriate for paper tensile strength?

[Mickyr]

Hi all, first time forum user,
We have a customer who insists on using Linearity and Bias to evaluate one of our instruments. The product being measured is paper and is a cross direction sample (20 data pts). The test is for Tensile strength. They are comparing our results with another manufacturer. I don't understand the application of Linearity for what is esentially a single point evaluation (data range is 85-90 lbf). Is this an appropriate use of this statistical tool? Can someone point me in the direction of understanding what these evaluations really represent?

Thanks for any assistance.

 

[Miner]

Re: When to use Linearity and Bias studie

A Linearity study will essentially verify the accuracy/bias of the instrument throughout its entire measurement range.

Your customer's request may indicate that they have experienced a problem with linearity in the past. This may be an opportunity for you.

If the customer is considering the purchase of one of your machines, they are well within their rights to request this, and are showing good judgement.

 

[Mickyr]

Thank you "Miner"
I was hoping to get a better understanding of the nuts and bolts of using these two evaluations. If I provided data could you help to explain the results?

 

[justncredible]

Tensile is a destructive test? A tear test? What do you use for a reference value?

Provide the data, and maybe I can throw up the results, but I not sure if they will have value.

 

[AndyN]

What your customer is going to establish by having you do this study is whether on not the machine pulls (and hence indicates) the test in a linear manner.

You say it's a single point evaluation but twice in your post you mention a range (20 data points) and 85 - 90 lbs. Hence the need to know if the machine's operation is linear - that is measured value vs indicated value over that range. It could be that if the machine response was plotted, you might see a variation all along the 85 - 90 lbs range and the study is there to quantify how much that variation is.

 

[Miner]

Thank you "Miner"
I was hoping to get a better understanding of the nuts and bolts of using these two evaluations. If I provided data could you help to explain the results?

We can look at it. If it was properly performed, we can do the analysis. If not...

 

[Statistical Steven]

It is not a linearity study, but rather a method comparison study. What they are assessing is if your method is "linear" and if any bias is present when compared to their method. They are assuming that their method is "correct" and want to see how much if any bias is in your method.

There are lots of issues with this approach, but with destructive tests you need to determine how to best set up such a test.