2 participants in inter-laboratory comparison tests

[EMengineer]

Good day,

I am stuck with interlaboratory comparison results statistical analysis. There are a few main methods, but they work when there are more than 3 participants.

The comparison tests object is power supply unit, thus in this case there is no reference value/material. Test methods would be power measurement, enclosure temperature, leakage currents.

Could someone give the reference to relevant statistical method when only two laboratories are participating?

Thank you in advance.

 

[Miner]

You should be able to use a Bland-Altman or Youden plot to analyze your study. BevD and I have multiple posts on these methods that you can search. See Bev D's resource on a sister forum.

 

[EMengineer]

Hi, thank you for your answer, as I've checked these methods are used when there are large amount of results.
In my case there would be only two results - one from our laboratory, another from our partners. For example we measure creepage distance, in this case we get average 2,5 mm, and other laboratory 2,4 mm and we need to perform statistical analysis for these results. If there are the third participant we could use Z-score method, but in our case there are only two participants. Is there any suitable method for this case? We need to determine the limits for positive results. I have found that in some cases total uncertainty of participants is used, for example - we have measured creepage distance with 3% uncertainty, other laboratory with 5%, so our results should not differ more than 8%.
Does anyone have more information about such case?

 

[tacom]

Perform at least 7 experiments in 2 samples in the same laboratory and check the results with F tets and t test.
Repeat this procedure in several different samples

"F" tets for variance compliance; "t" test for arithmetic mean compliance