Show solutionHide solution
Question 8
More repetitions reduce one uncertainty component
A measurement method has a known random standard deviation g per reading. Assume independent readings of a stable sample with the same distribution. The standard uncertainty of their mean is . An uncorrected, constant calibration offset of g affects every reading.
Diagram shown in the original SVG worksheet.
Calculate for and .
How many readings are needed to reduce this random component to 0.01 g or less under the model?
Does collecting 36 readings remove the calibration offset? Is a guaranteed interval?
Question 8 - Solution
More repetitions reduce one uncertainty component
Diagram shown in the original SVG worksheet.
(a) Apply the supplied model
Nine times as many readings reduce this component by a factor of three, not nine.
(b) Find the required sample count
This conclusion depends on the stated independence and stability assumptions.
(c) Separate random spread from offset
The constant g offset remains in the mean regardless of ; it requires calibration correction. A standard uncertainty describes a statistical spread, not a guaranteed bound. Without an additional distributional model, no specific coverage probability follows for .
Check and interpretation
At , the offset is four times the random standard uncertainty. More readings alone cannot remove it.