# How can I calculate uncertainty based on measurements from multiple separate runs?

I am using standard test blocks to find the uncertainty of a supermicrometer, which is capable of measuring down to 1/10 of a micron. The issue is that it measures the EXACT same length within a run of even 100 measurements. Understandably, it is suspicious to calculate uncertainty based on data with zero variation; however, the value that it consistently measures is different every time I zero the supermicrometer at a different position. For example, I may get a value of 5.0000 mm repeatedly during one run, zero the supermicrometer, and then get a value of 5.0003 mm repeatedly during the next run.

So the question is, how can I take into consideration variation between runs? Traditionally, one would use a single run to find uncertainty, but I don't think I can do that.

• Has the temoerature changed between each run? For small differences even a small variation of temperature can have an effect. – Solar Mike Sep 1 '18 at 5:41
• What happens if you alternate between measuring two different lengths, without recalibrating? (e.g. 5mm, 6mm, 5mm, 6mm, etc). – alephzero Sep 1 '18 at 7:46
• The temperature was the same for all runs. I have not tried alternating between two different lengths. The supermicrometer was re-zeroed after each run, NOT recalibrated. – Gevorg Ter-Isahakyan Sep 1 '18 at 15:07
• I recently attempted to alternate between two lengths. It did not make a difference. – Gevorg Ter-Isahakyan Sep 6 '18 at 12:50