- Foundation
- About 20 min
- 5 practice questions
Mean deviation for grouped data
Find the mean deviation about the mean and about the median for discrete and continuous frequency distributions, and state the limitations of mean deviation.
If something looks wrong, please tell us.
Large data sets are usually summarised in frequency tables. The idea of mean deviation stays the same, the average distance from the centre, but each value now counts as many times as its frequency, and for class intervals we use mid-points.
Discrete frequency distributions
Let the values x₁, …, xₙ occur with frequencies f₁, …, fₙ, and let N = Σfᵢ. Then:
x̄ = Σfᵢxᵢ / N, M.D.(x̄) = Σfᵢ|xᵢ − x̄| / N and M.D.(M) = Σfᵢ|xᵢ − M| / N.
For the median of a discrete distribution, arrange the values in increasing order, form cumulative frequencies (c.f.), and take the value whose cumulative frequency first equals or exceeds N/2.
Worked example 1: about the mean
Made-up data: the number of books read in a month by 20 students.
| xᵢ | fᵢ | fᵢxᵢ | |xᵢ − 5.5| | fᵢ|xᵢ − 5.5| |
|---|---|---|---|---|
| 2 | 4 | 8 | 3.5 | 14 |
| 4 | 5 | 20 | 1.5 | 7.5 |
| 6 | 5 | 30 | 0.5 | 2.5 |
| 8 | 4 | 32 | 2.5 | 10 |
| 10 | 2 | 20 | 4.5 | 9 |
| Total | 20 | 110 | 43 |
M.D.(x̄) = 43 ÷ 20 = 2.15.
Worked example 2: about the median, same data
Cumulative frequencies are 4, 9, 14, 18, 20 and N/2 = 10. The first cumulative frequency at least 10 is 14, belonging to x = 6, so M = 6. The absolute deviations |xᵢ − 6| are 4, 2, 0, 2, 4, and Σfᵢ|xᵢ − 6| = 16 + 10 + 0 + 8 + 8 = 42. M.D.(M) = 42 ÷ 20 = 2.1, slightly less than the mean deviation about the mean. (In general the sum of absolute deviations is smallest when taken about the median.)
Continuous frequency distributions
For class intervals, assume each class is represented by its mid-point xᵢ and use the same formulae. The median of a continuous distribution is found from the median class (the class whose cumulative frequency first reaches N/2):
Median M = l + [(N/2 − C)/f] × h, where l = lower limit of the median class, f = its frequency, h = class width and C = cumulative frequency of the class before it.
Mean deviation for a continuous distribution
- Mid-points xᵢ = (lower + upper limit)/2
- Centre x̄ = Σfᵢxᵢ/N, or M from the median class
- Absolute deviations |xᵢ − x̄| or |xᵢ − M| for each mid-point
- Weight and average Σfᵢ|xᵢ − a| / N
Worked example 3: about the mean
Made-up marks of 40 students:
| Class | fᵢ | Mid-point xᵢ | fᵢxᵢ | |xᵢ − 24.5| | fᵢ|xᵢ − 24.5| |
|---|---|---|---|---|---|
| 0-10 | 5 | 5 | 25 | 19.5 | 97.5 |
| 10-20 | 8 | 15 | 120 | 9.5 | 76 |
| 20-30 | 15 | 25 | 375 | 0.5 | 7.5 |
| 30-40 | 8 | 35 | 280 | 10.5 | 84 |
| 40-50 | 4 | 45 | 180 | 20.5 | 82 |
| Total | 40 | 980 | 347 |
M.D.(x̄) = 347 ÷ 40 = 8.675 marks. Short cut for the mean: with assumed mean A = 25 and class width h = 10, the step deviations dᵢ = (xᵢ − 25)/10 are −2, −1, 0, 1, 2, Σfᵢdᵢ = −10 − 8 + 0 + 8 + 8 = −2, and x̄ = A + h × Σfᵢdᵢ/N = 25 − 0.5 = 24.5.
Worked example 4: about the median
Cumulative frequencies: 5, 13, 28, 36, 40; N/2 = 20, so the median class is 20-30 with l = 20, C = 13, f = 15, h = 10. M = 20 + (20 − 13)/15 × 10 = 20 + 14/3 ≈ 24.67. The absolute deviations of the mid-points from 74/3 are 59/3, 29/3, 1/3, 31/3, 61/3; weighting by the frequencies gives a total of 1034/3 ≈ 344.67. M.D.(M) ≈ 344.67 ÷ 40 ≈ 8.62 marks, again a little below the mean deviation about the mean.
Limitations of mean deviation
- When the data are very variable, the median may not represent them well, so the mean deviation about the median is not fully reliable.
- The sum of absolute deviations about the mean is more than that about the median, so the mean deviation about the mean is not the best possible measure.
- Absolute values are awkward in algebra, so mean deviation cannot easily be used in further calculations. This motivates the standard deviation.
Key takeaways
- Grouped data: weight each absolute deviation by its frequency and divide by N.
- Continuous classes: use mid-points; find the median from the median class.
- Mean deviation is easy to understand but hard to handle algebraically.
Check your understanding
-
1. For the distribution x: 1, 3, 5 with frequencies 2, 3, 5, what is the mean?
- 3
- 3.6
- 3.4
- 9
Show answer
Answer: B. Σfx = 2 + 9 + 25 = 36 and N = 10, so the mean is 36/10 = 3.6.
-
2. For x: 1, 3, 5 with f: 2, 3, 4, what is the mean deviation about the median?
- 4/3 ≈ 1.33
- 1.4
- 0.8
- 2
Show answer
Answer: A. N = 9 and N/2 = 4.5. Cumulative frequencies are 2, 5, 9; the first one at least 4.5 is 5, at x = 3, so M = 3. Σf|x − 3| = 2(2) + 3(0) + 4(2) = 12, and M.D. = 12/9 = 4/3 ≈ 1.33.
-
3. In a continuous distribution, N = 50 and the median class is 40-50 with frequency 10. The cumulative frequency before it is 20. What is the median?
- 45
- 40
- 50
- 43
Show answer
Answer: A. M = l + [(N/2 − C)/f] × h = 40 + [(25 − 20)/10] × 10 = 40 + 5 = 45.
-
4. For a continuous distribution, which value represents the class 30-40 in the mean deviation calculation?
- 30
- 40
- 10
- 35
Show answer
Answer: D. Each class is represented by its mid-point (30 + 40)/2 = 35.
-
5. Assertion (A): Mean deviation is not well suited to further algebraic work. Reason (R): It is based on absolute values of deviations.
- Both A and R are true, and R is the correct explanation of A
- Both A and R are true, but R is not the correct explanation of A
- A is true, but R is false
- A is false, but R is true
Show answer
Answer: A. Both are true. Absolute values are difficult to manipulate algebraically, which is exactly why mean deviation is hard to use in further theory. R explains A.
Sign in to keep track of the lessons you finish.