Errors in Measurement: The Science of Being Wrong Correctly
JEE/NEET Physics · Units & Measurements series · Part 4 of 6 · All parts →
- Absolute error = |your value − true value|; it carries the unit
- Relative error = absolute ÷ true value; fractional, unitless
- Percentage error = relative × 100
- In sums/differences: ABSOLUTE errors add; in products/quotients: RELATIVE errors add
- In powers: relative error multiplies by the exponent (A³ → ×3)
- Mean absolute error and ‘±’ notation: 20.1 ± 0.3 cm is a promise, not a guess
Every measurement ever made is wrong — the only question is by how much, and whether you KNOW by how much. Error analysis is the science of being wrong correctly. In Part 3 you learned how significant figures control the honesty of a single written number. This part goes deeper: once you have made several imperfect measurements, how do you report them honestly — and what happens to that honesty when you add, multiply, or raise them to powers? Welcome to Part 4 of the Units & Measurements series.
- Why error analysis matters at all
- Absolute, relative, percentage: the three reporting levels
- How errors combine in arithmetic
- The mean-absolute-error protocol
- Solved examples
- Common mistakes
- This physics in your daily life
- Frequently asked questions
- Practice set
- Recap
Why Error Analysis Matters — No Measurement Is Perfect
Before the formulas, fix one idea in your mind: error does not mean mistake. In physics, “error” is the unavoidable gap between a measured value and the true value. Even the finest scientist on Earth, using the best instrument money can buy, cannot measure anything exactly. Instruments have limits (least counts), observers have reaction times, and the very act of measuring disturbs what is being measured.
What separates good science from bad science is not whether errors exist — they always do — but whether the experimenter tracks, quantifies, and reports them. A value of 20.1 cm with an honest “± 0.2 cm” attached is more scientifically valuable than a bare “20.1 cm” with no stated uncertainty, because only the first can be compared with other results, trusted in further calculations, or improved upon. Error analysis, then, is not bookkeeping — it is the licence that lets physics build on itself.
Absolute, Relative, Percentage: The Three Reporting Levels
Measure a rod four times: 20.1, 20.4, 19.9, 20.0 cm. No single reading is “the answer” — repeated readings scatter because of small, unpredictable fluctuations. The best estimate of the true value is the arithmetic mean of all readings:
Mean = (20.1 + 20.4 + 19.9 + 20.0)/4 = 20.1 cm.
Now each reading is compared with this mean. Each reading’s absolute error is its distance from the mean, always written positive:
|20.1 − 20.1| = 0.0, |20.4 − 20.1| = 0.3, |19.9 − 20.1| = 0.2, |20.0 − 20.1| = 0.1.
Average those deviations and you get the mean absolute error: (0.0 + 0.3 + 0.2 + 0.1)/4 ≈ 0.15 ≈ 0.2 cm. The complete, honest report becomes 20.1 ± 0.2 cm — a claim that the true length lies somewhere in the band from 19.9 to 20.3 cm.
But notice something: an error of 0.2 cm on a 20 cm rod is trivial; the same 0.2 cm error on a 2 cm rod would be serious. Absolute error alone hides this. That is why we climb the reporting ladder:
| Level | Definition | Unit? | What it tells you |
|---|---|---|---|
| Absolute error Δa | |measured − true| | same unit as the quantity | the raw size of the gap |
| Relative error Δa/a | absolute ÷ true value | none — a pure ratio | how big the gap is compared to the quantity itself |
| Percentage error | relative × 100 | % | the same honesty in everyday language |
For the rod: relative error = 0.2/20.1 ≈ 0.01, so percentage error ≈ 1%. One measurement, three levels of reporting — and each level exists because a different question demands it.
How Errors Combine in Arithmetic
Measured quantities are rarely used alone; we add them, multiply them, square them. Every operation transfers uncertainty from the inputs to the output. The rules are few and absolute:
| Operation | What adds | Why |
|---|---|---|
| A + B or A − B | absolute errors | the ± bands stack — one bad band shifts the sum by exactly that much |
| A × B or A ÷ B | relative errors | percentage slop multiplies |
| Aⁿ | relative error × n | exponentiation repeats the slop n times |
| Z = A²B³/C | 2(ΔA/A) + 3(ΔB/B) + (ΔC/C) | the general power rule, all at once |
The asymmetry trips up every beginner: addition pairs with absolute errors, multiplication pairs with relative errors. Why? When you add A ± ΔA to B ± ΔB, the worst case is a shift of ΔA + ΔB in the same direction — the bands stack in units. But when you multiply, a 2% stretch in A stretches the product by 2% regardless of B; percentage slop passes through multiplication untouched, so the percentages from each factor simply compound.
One more subtlety worth committing to memory: for subtraction, absolute errors still add — never subtract. If A = 10.0 ± 0.1 and B = 9.9 ± 0.1, then A − B = 0.1 ± 0.2: the uncertainty band can be larger than the answer itself. This is why experiments that hinge on the difference of two nearly equal numbers are notoriously hard.
| Letter | What it means (plain words) | Value / unit |
|---|---|---|
| ΔA | the absolute error in A | same unit as A |
| ΔA/A | relative error — the honest fraction | unitless (a pure ratio) |
| ± x | the uncertainty band of the result | truth promised within x |
The Mean-Absolute-Error Protocol
Whenever you take several readings of the same quantity, follow this four-step protocol — it is the standard expected in lab exams and viva questions:
Step 1 — Take n readings of the quantity under identical conditions.
Step 2 — Average them to get the best estimate of the true value.
Step 3 — Compute each reading’s absolute deviation from that mean (always positive).
Step 4 — Average the deviations to get the mean absolute error, then report:
Result = mean ± mean absolute error.
The ‘±’ is the crucial part. It is a promise about where the truth lives, not a confession of sloppiness. That promise is why the ± notation appears on every lab report, medicine label, and election poll in the world — a number without a band cannot be checked against anything, and unverifiable numbers are not science.
Solved Examples
Step 1: Absolute error = |20.4 − 20.0| = 0.4 cm.
Step 2: Relative error = 0.4/20.0 = 0.02.
Step 3: Percentage error = 0.02 × 100 → percentage error = 2%. ✔
Answer: 2%
Step 1: Relative error in T = 0.1/2.0 = 5%.
Step 2: T² doubles it: 2 × 5% = 10%.
This is exactly why period-squared graphs stretch errors — and why examiners love this question. ✔
Answer: 10%
Step 1: Value: P = 4.0 × 2.0 = 8.0.
Step 2: Relative errors: 0.1/4.0 = 2.5%; 0.2/2.0 = 10% → total 12.5% (product rule).
Step 3: ΔP = 12.5% of 8.0 = 1.0.
P = 8.0 ± 1.0 — one honest band around the answer. ✔
Answer: P = 8.0 ± 1.0
- Adding relative errors in a subtraction. A − B still adds ABSOLUTE errors — subtraction of values never cancels the slop.
- Forgetting the exponent multiplier. In A³, relative error ×3, not ×1 — the single most-tested slip.
- Averaging before subtracting to reduce error. Errors never subtract; uncertainty only grows through arithmetic, never shrinks. Repeating readings narrows the error of the mean itself, but any calculation built on measured values only accumulates uncertainty.
- Reporting 20.145 ± 0.2. The error digit and the last value digit must align — write 20.1 ± 0.2. The uncertainty dictates where your value stops.
- Treating absolute and relative errors as interchangeable. Before combining anything, ask: is this operation addition-like or multiplication-like? Then choose the matching error type.
This Physics in Your Daily Life
- Election polls ‘48% ± 3%’ — the ± band you see on TV is exactly this chapter: a promise about where the truth lives.
- Medicine doses (‘500 mg ± tolerance’) — pharmaceutical manufacturing lives inside error budgets; pills must deliver the promised digit.
- Cricket ball-tracking (LBW ‘umpire’s call’) — projected paths carry uncertainty ellipses; the third umpire rules WITHIN the error band.
- Your phone GPS says ‘accuracy ~5 m’ — the chip reports its own ΔA; navigation apps plan routes through error ellipses, not points.
- Weather ’30 ± a few °C’ — ensembles are literally error arithmetic: many forecasts averaged, spread reported honestly.
- Fuel-efficiency figures (’18 km/l, tested under standard conditions’) — every certified number hides a tolerance band decided by error analysis.
A blind archer fires arrows. Where they land is each measurement; the cluster’s centre is the mean; the scatter size is the error. Reporting a value without its scatter is showing the bullseye but hiding the spread — technically true, practically a lie. The ± is the archer’s honest confession.
Weigh a coin 4 times: 10.1, 9.9, 10.2, 9.8 g → mean 10.0, mean absolute error ≈ 0.15 g. Report 10.0 ± 0.2 g. Multiply two such measurements and the relative slops add: (0.2/10) + (0.2/10) = 4% total — the honest band on the product is wider than on either factor. Arithmetic amplifies dishonesty unless you track it.
Draw each measurement as a dot on a number line with a shaded halo around it (the ± band). Stack two quantities: the halos stack too (addition). Multiply: the halos’ percentage widths stack. Every arithmetic step is a new picture with a fatter halo — the picture never lies about the slop growing.
Frequently Asked Questions
Q. Is “error” the same as “mistake”?
No. A mistake (misreading a scale, recording the wrong digit) is avoidable human error and should be corrected or discarded. A measurement error is the unavoidable, quantifiable gap between measured and true values — it is managed, not eliminated.
Q. Can percentage error be negative?
Not in JEE/NEET convention. Errors are magnitudes of deviation, always taken positive. A negative sign in intermediate work only indicates direction, and is dropped in the final error.
Q. Why do we take the mean of several readings?
Random fluctuations push individual readings above and below the true value roughly equally, so averaging cancels much of the scatter. The mean is the best single estimate, and the spread of deviations tells you how much to trust it.
Q. Why does subtraction make errors worse, not better?
Because the ± bands from both quantities stack even as the values cancel. A − B can be tiny while ΔA + ΔB stays large, so the relative error of a difference can explode.
Q. How many decimal places should the error carry?
Usually one significant figure. Round the error first, then round the value so its last digit sits at the same place as the error’s last digit: 20.145 ± 0.2 becomes 20.1 ± 0.2.
Practice set (answers hidden — try first)
(NEET-level) True 40.0, measured 41.2. % error:
(JEE Main-level) % error in V = 4/3 πr³ if r has 2% error:
(NEET-level) Errors in Z = A + B combine as:
(Concept) Readings 10.1, 10.2, 10.0. Mean absolute error:
(JEE Main-level) Z = A²/B; % errors 2% and 1%:
- absolute error |measured − true|, in units
- relative = absolute ÷ value; percentage × 100
- + − : absolute errors add; × ÷ : relative errors add
- Aⁿ multiplies relative error by n
- report as mean ± mean absolute error
- 🔁 absolute/relative/percentage error ladder
- 🔁 ± rule for sums vs products
- 🔁 power rule: Aⁿ → n × relative error
- 🧠 Chant: ‘plus/minus: absolutes; times/divide: relatives’.
- 🧠 Powers: ‘squared doubles the %, cubed triples it’.
- 🧠 Alignment: the error digit fixes where the value’s digits stop.
- 🏠 Daily: ‘48% ± 3%’ polls = this chapter on TV.
- 🏠 Daily: GPS ‘accuracy ~5 m’ = the ± in your pocket.
Quick revision
- Absolute error = |your value − true value|; it carries the unit
- Relative error = absolute ÷ true value; fractional, unitless
- Percentage error = relative × 100
- In sums/differences: ABSOLUTE errors add; in products/quotients: RELATIVE errors add
- In powers: relative error multiplies by the exponent (A³ → ×3)
- Mean absolute error and ‘±’ notation: 20.1 ± 0.3 cm is a promise, not a guess
Have a doubt on this topic?
