Sample Size, Significant Figures, Scale



One of our locations was recently cited during a Third Party audit for not establishing an appropriate sample size for calibration.

It may be that I've lost something in translation (I'm American, and my linguistic skills are very limited) - this is a finding from our Chinese facility.

They are asking:
What is the appropriate sample size for calibrating a measuring device with 0.0000 or 0.00000 decimal places of precision?

Frankly, I thought Calibration was a check, and that all the sampling sizes and everything came into play when establishing device uncertainty.

Next, I am utterly ignorant about the 0.0000 vs 0.00000 thing. I mean, I'm not approaching quantum behaviors. This is simply good old-fashioned, completely stable Microns.

Does anyone have any insight that may help?
I've been through ISO10012 and ISO5725. I may have missed it, but I didn't see Decimal Places called out for Calibration Sample Sizes?

Any help is appreciated.

(by the way, the Cove is excellent, I am glad it was able to return. And, i apologize for being a shy poster. This place has helped me a lot over the years)


Trusted Information Resource
Let me wade into these murky waters.

First, what were you being audited to? ISO 17025? FDA? A customer's requirements? If it was to a standard could you find out chapter and verse of what was being cited so we can check to see what they mean.

As for sample size based on resolution, typically you are calibrating to manufacturer specifications (unless otherwise directed). If the manufacturer calls out for a certain number of readings, or measuring at a given average setting, etc. then that is what you need to do. If there is no such guidance the instrument should be expected to report the correct value, within tolerance and uncertainty, for each reading.


dw -

Aw, dangit! You are making me go find some facts. I was hoping for the easy solution. I'm afraid if I lift this stone, the snake will bite me.

It was for FDA preparedness, not ISO17025, and not specific to a Customer's requirement. I would have been able to tell them ask the customer for guidance, had it been as easy as that.

I'll go ask for more facts. (And pull on my hip waders and my anti-snake bite boots - I'm afraid it is about to get deep and dangerous.)


Trusted Information Resource
The "sample size" flashed me back to FDA "requirements". Here is a link to a real good article explaining what they are probably looking for:

A couple of the relevant points for your case are:
"Calibration should include at least three points across the full range of the instrument to help ensure linearity. Calibration experts often recommend five points: the more points, the more assurance of accuracy."
"Prepared procedures must determine when to challenge repeatability (precision) for each test point where repeatability errors can produce substantial errors (e.g., mechanical pressure indicators)."

I've run into customers who have required multiple measurements on instruments that do not have repeatability tolerances (like DMMs) that are not lumped into the overall instrument specification. In our case we just would repeat 3 measurements and verified they were all in tolerance, that seemed to make them happy!
Top Bottom