Measurement · Lesson 2 of 2
Uncertainty
Random errors cause scatter between readings; systematic errors shift results consistently. Repeats and averaging reduce random effects but do not fix an incorrect zero. Two results are equal within experimental accuracy if they differ by no more than 10%.
By the end, you can
- Identify random and systematic errors.
- Repeat measurements.
- Calculate averages.
- Evaluate limitations.
- Use the terms true value, measurement error, accuracy and precision.
- Use the terms repeatability, reproducibility and validity.
- Decide whether two results are equal within the limits of experimental accuracy, taken as ±10%.
Experimental uncertainty
Every measurement is affected by errors, and they come in two kinds. Random errors make repeated readings scatter unpredictably either side of the true value: reaction time when starting and stopping a stopwatch, judging the moment a ball passes a mark, a draught disturbing a balance. Systematic errors shift every reading by the same amount in the same direction: a balance that has not been zeroed, a ruler always read from the same wrong angle, a stopwatch that runs slow.
Random errors are reduced by repeating each measurement and calculating the mean, or average: add the readings and divide by how many there are. Times of 2.31 s, 2.28 s and 2.34 s give a mean of (2.31 + 2.28 + 2.34) ÷ 3 = 6.93 ÷ 3 = 2.31 s, and the scatter of the single readings partly cancels out. Timing twenty swings of a pendulum and dividing by twenty works in the same way, because the reaction-time error is shared out over all the swings.
Before calculating a mean, look for an anomalous result: a reading far outside the pattern of the others, such as 2.85 s among readings close to 2.3 s. Check it or repeat it if you can, and leave it out of the mean rather than let one bad reading drag the answer. Averaging does nothing about a systematic error, because every reading is shifted the same way; the error has to be found and removed, by checking the zero, reading scales at eye level or using a better instrument.
Evaluating the limitations of an experiment means identifying what stopped the results being better and saying how it could be improved. Be specific: "the stopwatch was started by hand, so reaction time affected each reading; use a light gate" earns credit, while "human error" names no problem and no fix. Consider too whether there were enough readings, and whether their range was wide enough to support the conclusion.
- mean = sum of the readings ÷ number of readings
- The average of a set of repeated readings, with any anomalous result left out.
Worth remembering
- Random errors scatter readings; systematic errors shift every reading the same way.
- Repeat readings and take the mean to reduce the effect of random errors.
- Leave anomalous results out of a mean; averaging cannot remove a systematic error.
- Name a specific limitation and a specific improvement, not "human error".
The language of measurement
Scientists use a set of words to describe measurements, and using them correctly makes an evaluation clear. The true value of a quantity is the value that would be obtained in an ideal measurement. The measurement error is the difference between a measured value and the true value. A measurement is accurate if it is close to the true value.
Precision describes how close measured values of a quantity are to each other: a set of readings that agree closely is precise, whether or not it is close to the true value. The same word is used for how finely an instrument can measure, which is why readings are recorded to the precision of the instrument. Darts tightly grouped in one corner of a dartboard are precise but not accurate; darts tightly grouped on the bullseye are both precise and accurate.
A measurement is repeatable if the same or a similar result is obtained when it is repeated under the same conditions, by the same method, within the same experiment. It is reproducible if the same or a similar result is obtained under different conditions, by a different method or in a different experiment, for example by another group with different apparatus. An experiment is valid if it tests what it says it will test: it must be a fair test in which only the independent and dependent variables change and the control variables are kept constant.
At this level, two results are taken to be equal within the limits of experimental accuracy if they differ by no more than 10%. To check, find the difference between them and compare it with 10% of one of the values. Two values of the acceleration of free fall, 9.4 m/s² and 9.8 m/s², differ by 0.4 m/s², which is less than 10% of 9.8 m/s² (0.98 m/s²), so they are equal within experimental accuracy. Values of 8.0 m/s² and 9.8 m/s² differ by 1.8 m/s², which is more than 10%, so they are not.
Worth remembering
- True value: the ideal result. Measurement error: the difference between a measured value and the true value.
- Accurate: close to the true value. Precise: measured values close to each other.
- Repeatable: same result by the same method; reproducible: same result another way; valid: a fair test.
- Two results within 10% of each other are equal within the limits of experimental accuracy.
Worked example
Readings 2.1, 2.3 and 2.2 s average to 2.2 s. A stopwatch offset of +0.2 s would still bias that mean.
Investigate scatter and systematic error
Change a value and watch the readings follow. Open it on the lab stage
Investigate scatter and systematic error
There is no zero offset, and the readings lie up to 0.484 cm from the best-fit line.
- Furthest point from the best-fit line
- 0.484 cm
- 3.13 cm − 2.646 cm at 1 s
- Best-fit intercept
- 1.165 cm
- 1 cm true + 0 cm offset + 0.165 cm random
- Best-fit gradient
- 1.965 cm/s
- (20.815 cm − 1.165 cm) ÷ 10 s
Goal
Find out how random scatter and a zero offset each change a set of readings, and which of them taking more repeats can fix.
Predict, where you are
If Ithe, thewill
Say what you expect before you move anything. Check makes the change for you.
Test, still to come
Change one thing at a time and watch the readings. The notebook keeps a reading each time you record one.
Explain, still to come
After you check a prediction, this says why it happened.
Try this · 0 of 3
- Turn the random scatter up to 1.5 cm with no zero offset. Does the best-fit line still follow the true line?, not yet
- Set a zero offset of −1 cm with no scatter. How far does every reading move?, not yet
- With 1 cm of scatter and a 1 cm zero offset, take 10 repeats at each time. Which error does averaging reduce, and which does it leave?, not yet
reading = true value + zero offset + random error
at 5 s: 11 cm + 0 cm − 0.229 cm = 10.771 cm
Investigation notebook · 0 readings
Change one variable at a time and record the results. These readings stay here until you leave this lesson or reset the notebook.
About this model
Made-up readings of a distance that truly grows as 2 cm/s × time + 1 cm. Each reading’s random error comes from a fixed set drawn once, up to the scatter either side, so the same settings always give the same readings. The best-fit line is the least-squares straight line through the plotted points.
Practice
Warm-up · not marked
Check your understanding
Answer in your own words, then compare with the worked answer. Nothing here is saved or scored.
Worked answer
Yes: they differ by 1 N/m, which is less than 10% of either value.
Check your method, units and explanation against it.
Marked practice
Set 1 of 5 · easiest first · Checked as you go, with the worked answer. Sign in or create a free account to save your progress.