Systematic Error, Controls and the Design of a Measurement
Repeating a measurement shrinks the random part and leaves the systematic part untouched, so the design has to attack systematics directly through calibration, null tests and blind analysis. Most disagreements between published results are arguments about the systematic budget.
What a learner can do afterwards
- Separates random from systematic contributions in a described experiment
- Proposes a null test or a blind analysis for a stated risk of bias
- Builds an error budget and identifies which term dominates
1 · Read
Every measurement carries two kinds of trouble. Random error scatters your readings around the truth, and repeating the measurement shrinks it. Systematic error is a built in offset that survives averaging: repeat a hundred times and the offset sits exactly where it started. Most disagreements between published results are arguments about the systematic budget, not about sample size.
A scale misses by 3 grams on a 180 gram standard. The absolute error is 3 grams, stated in units. The relative error divides by the expected value, giving about 1.7 percent. An error budget lists every uncertainty source with its size, and the dominant term decides the answer: you attack the biggest entry first, because shrinking small ones barely moves the total.
Good design attacks systematics directly instead of hoping they average away. A null test is a measurement run where the signal must read zero, so a nonzero answer exposes a systematic offset on the spot. Blind analysis hides the group labels until the end, which stops expectation from steering judgment. Calibration against a standard and a control group serve the same purpose: they give bias nowhere to hide.
Before you measure anything, design the check that would expose a systematic error rather than average it away. Write down the null test, the control, and the blinding step in advance. If a result depends on one untested assumption, that assumption is your dominant term whether the budget admits it or not.
Repeats shrink random scatter, while only calibration, null tests, and blinding expose systematic bias.
2 · Watch
Take it off screen
Where it sits
Learn first
Where this leads
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.