Measurement Error

[Watts]

Main Question or Discussion Point

Could some please distinguish the difference between measurement error and measurement uncertainty. I have actually held conversations at conferences over this issue with statisticians from NIST and no one has ever been able to give me a consistent answer. It seems to be more philosophical than anything. I was just wondering if any body could shed some thoughts on this subject?

 

[EnumaElish]

See Uncertainty. My "educated guess" is that uncertainty is minimized when, say, all error terms (in the sense of residual = measured value - true value) are either -1 or +1. But error can further be minimized when all error terms are randomly distributed, say, between -0.01 and +0.01, in which case uncertainty may be greater than in the previous (binary) case because eror terms are "all over" (albeit within a tiny interval).

 

[balakrishnan_v]

The former is mean-actual value and the latter is the variance

 

[Watts]

Thanks

That helps matters. The statement made in the link ("Because of an unfortunate use of terminology in systems analysis discourse, the word "uncertainty" has both a precise technical meaning and its loose natural meaning of an event or situation that is not certain.") provided still leaves things open for discussion. I will study the information provided and the link information further. On a second note it is basically another way of measuring dispersion.

 

[EnumaElish]

The former is mean-actual value and the latter is the variance

You are saying that uncertainty is identical to + standard deviation; is that correct?