• Foundation
  • About 20 min
  • 5 practice questions

Tally marks, loss of information and unequal classes

Construct a frequency distribution from raw data using tally marks, explain the loss of information in grouping, and see why unequal classes can represent data better.

If something looks wrong, please tell us.

By the end of this lesson you can

  • Use tally marks to count class frequencies from raw data.
  • Explain and measure the loss of information when data are grouped.
  • Explain how splitting crowded classes into narrower ones gives more representative class marks.

The fifth question in building a frequency distribution is how to get each class's frequency. The frequency of an observation is the number of times it occurs in the raw data; the class frequency is the number of observations falling in a class. Class frequencies are counted with tally marks.

Tally marking

Go through the raw data once. For each observation, put a tally (a short stroke) against the class it belongs to. To make counting easy, tallies are grouped in fives: four strokes, then a fifth stroke drawn across them. Below, a crossed group of five is written as ||||/ (read: five), so ||||/ || means 7. The class frequency is the number of tallies against the class.

Worked example 1: building a distribution from raw data

Illustrative raw data: marks out of 50 of 30 made-up students. 12, 27, 35, 8, 41, 22, 19, 30, 25, 33, 47, 15, 28, 36, 21, 10, 39, 26, 44, 31, 18, 24, 29, 38, 20, 34, 6, 27, 32, 23. Use exclusive classes 0-10, 10-20, ..., 40-50 with the upper limit excluded.

Illustrative data: tally marking
Class (marks)ObservationsTallyFrequencyClass mark
0-108, 6||25
10-2012, 19, 15, 10, 18||||/515
20-3027, 22, 25, 28, 21, 26, 24, 29, 20, 27, 23||||/ ||||/ |1125
30-4035, 30, 33, 36, 39, 31, 38, 34, 32||||/ ||||935
40-5041, 47, 44|||345
Total30

Check: 2 + 5 + 11 + 9 + 3 = 30, the number of observations. Note that 10 went in 10-20, 20 in 20-30 and 30 in 30-40, by the upper-limit-excluded rule. The maximum concentration is in 20-30: 11 ÷ 30 × 100 ≈ 36.7 per cent of students. Students scoring 20 to 40: 11 + 9 = 20, which is 20 ÷ 30 × 100 ≈ 66.7 per cent.

Loss of information

A frequency distribution summarises raw data and makes it concise, but it has an inherent shortcoming: it does not show the details of the raw data. After grouping, an individual observation has no significance in further calculations; every value in a class is treated as equal to the class mark. This is called loss of information. Much is gained in clarity, which more than makes up for it, but we should know what is lost.

Worked example 2: measuring the loss

From the table above, class 40-50 contains 41, 47 and 44. Actual total: 41 + 47 + 44 = 132; actual mean = 132 ÷ 3 = 44. Using the class mark: 3 × 45 = 135; implied mean 45. The class mark overstates by 1 mark per student. For class 0-10 (values 8 and 6), actual total = 14, actual mean = 7, but class mark total = 2 × 5 = 10, so the class mark understates by 2 marks per student. Small differences like these, added across classes, are the price of grouping.

Grouping makes data concise but causes loss of information: the frequency tells how many observations a class has, not what they are, and all of them are represented by the class mark.

Frequency distribution with unequal classes

When most observations crowd into a few classes, those classes may be split into narrower ones, so that each class mark lies closer to the values around which the observations concentrate. Sparse classes keep their wider width. The new class marks are then more representative of their data.

Worked example 3: splitting a crowded class

In Worked example 1, class 20-30 holds 11 of the 30 observations. Split it into 20-25 and 25-30. 20-25: 22, 21, 24, 20, 23 (5 values; class mark (20 + 25) ÷ 2 = 22.5). 25-30: 27, 25, 28, 26, 29, 27 (6 values; class mark 27.5). Check: 5 + 6 = 11. How representative? Actual mean of 20-25 values = (22 + 21 + 24 + 20 + 23) ÷ 5 = 110 ÷ 5 = 22, close to 22.5. Actual mean of 25-30 values = (27 + 25 + 28 + 26 + 29 + 27) ÷ 6 = 162 ÷ 6 = 27, close to 27.5. With the single class 20-30, all 11 values were represented by 25, which is 3 away from 22 and 2 away from 27.

From raw data to a frequency distribution

  1. Raw data Observations in no order
  2. Decide classes Equal or unequal; number, size and limits
  3. Tally One stroke per observation, in groups of five
  4. Count frequencies Check that they add to the total
  5. Class marks Represent each class in later calculations

Key takeaways: class frequencies are counted with tally marks grouped in fives, and the frequencies must add up to the number of observations. Grouping causes loss of information because every value in a class is treated as the class mark. Where observations crowd into a few classes, narrower (unequal) classes give class marks that represent the data better.

Check your understanding

  1. 1. Illustrative: the tally ||||/ ||||/ ||| stands for a class frequency of:

    1. 13
    2. 10
    3. 3
    4. 15
    Show answer

    Answer: A. Each crossed group ||||/ counts as 5. Two groups make 10, and three single strokes add 3, so 10 + 3 = 13.

  2. 2. Illustrative: a class 60-70 contains the values 61, 63 and 62. If the class mark is used for all three, the total is overstated by:

    1. 6
    2. 9
    3. 195
    4. 186
    Show answer

    Answer: B. Actual total = 61 + 63 + 62 = 186. Total using the class mark = 3 × 65 = 195. Overstatement = 195 − 186 = 9.

  3. 3. 'Loss of information' in a frequency distribution means that:

    1. some observations are thrown away
    2. the total frequency is reduced
    3. class limits become unclear
    4. the actual values within a class are not shown; all are taken as the class mark
    Show answer

    Answer: D. No observation is discarded and the total is unchanged; what is lost is the detail of each value, since every value in a class is represented by the class mark.

  4. 4. Narrower classes are formed in the crowded part of a distribution so that:

    1. the number of observations rises
    2. the data become qualitative
    3. class marks are closer to the values around which observations concentrate
    4. tally marks are not needed
    Show answer

    Answer: C. Splitting crowded classes makes each class mark more representative of the observations in it, which is the reason the chapter gives for unequal classes.

  5. 5. Assertion (A): The sum of class frequencies must equal the total number of observations. Reason (R): Every observation is tallied in exactly one class.

    1. Both A and R are true, and R is the correct explanation of A
    2. Both A and R are true, but R is not the correct explanation of A
    3. A is true, but R is false
    4. A is false, but R is true
    Show answer

    Answer: A. Both are true, and R explains A: because each observation is counted once and only once, adding all class frequencies gives the total number of observations.

Privacy · Terms · Refunds · Delivery