Did you know that a chart can contain more pupils than were surveyed, even though every addition on the page is correct? Imagine 30 pupils reporting their ages and a classmate creating two groups: “10–12 years” and “12–14 years.” Eight pupils are exactly 12. If the group labels include both endpoints, those eight can appear twice. The wrong total comes from overlapping categories, not difficult arithmetic.
Quick answer: a frequency-table class interval is a clearly defined range used to assign observations. For a table intended to count each person once, the age groups must be mutually exclusive (no value belongs to two groups) and should cover all the valid values being summarised. Replace ambiguous ranges such as “10–12” and “12–14” with non-overlapping integer-age groups such as “10–11”, “12–13”, “14–15”, or explicitly defined intervals such as 10 ≤ age < 12 and 12 ≤ age < 14.
- Check whether ages are recorded as completed years or more precise measurements.
- Specify which endpoints belong to which groups.
- Test every shared boundary, particularly 12 in the original example.
- Count each valid pupil exactly once, and record missing age responses separately.
- Add the group frequencies and compare them with the number of valid observations.
- Choose the correct denominator before calculating percentages.
- Keep the original age records long enough to verify the grouping, with appropriate privacy protection.
The diagnostic question: which group owns the twelve-year-old?
Write two headings: Ages 10–12 and Ages 12–14. Ask a child: “Where does age 12 go?” If they say “both”, the headings overlap. If they say “it depends”, ask them to write the rule that resolves the question. If they can identify that the same observation should not be counted in two exclusive classes, they understand the core issue. If they simply say “we should subtract the extra people at the end”, they are treating the symptom rather than repairing the classification.
A frequency distribution is a summary of data by category or interval. Each observation goes into a specified class and contributes one count in a single-choice table. For context, the US National Institute of Standards and Technology’s documentation on binning explains that binning computes frequencies for classes. A learner need not know statistical software to use the simple principle: decide exactly where a value belongs before counting it.
Fully worked example: thirty invented pupil ages
This is a fictional teaching dataset, not a real school record. Suppose 30 valid responses contain the following completed ages: 10-year-olds: 4; 11-year-olds: 6; 12-year-olds: 8; 13-year-olds: 7; 14-year-olds: 5. The original total is 4 + 6 + 8 + 7 + 5 = 30 pupils. All percentages and grouped totals must be checked against that 30.
A learner uses inclusive labels “10–12” and “12–14”. Group one counts ages 10, 11 and 12: 4 + 6 + 8 = 18. Group two counts ages 12, 13 and 14: 8 + 7 + 5 = 20. Their group totals add to 38, even though the sample contains 30. The excess is precisely the eight pupils aged 12: 38 − 30 = 8. This is the numerical signature of double counting in this particular example.
The tempting “fix” is to change the final total from 38 to 30 by subtraction while leaving both group counts unchanged. That makes the bottom line look correct but leaves the chart internally inconsistent. An 18-and-20 table still suggests two bars whose combined frequencies exceed the actual number of distinct pupils. The categories themselves must be rewritten.
Repair the table, not merely its bottom row
Because the ages are completed whole years, use the non-overlapping groups 10–11, 12–13 and 14–15, treating both listed integer endpoints as included. The first group contains 4 + 6 = 10 pupils. The second contains 8 + 7 = 15. The third contains the five pupils aged 14 and zero aged 15, so its count is 5. The corrected total is 10 + 15 + 5 = 30.
For this table, the three frequency percentages are 10/30 × 100 = 33⅓%, 15/30 × 100 = 50%, and 5/30 × 100 = 16⅔%. Their exact sum is 100%. If presented to one decimal place, they become 33.3%, 50.0% and 16.7%, which also sum to 100.0% here. Rounding can cause minor total discrepancies in other datasets, so the exact counts remain the best check.
Another valid method uses half-open intervals: 10 ≤ age < 12; 12 ≤ age < 14; 14 ≤ age < 16. These mean the left boundary is included while the right boundary is not. A pupil with recorded age exactly 12 goes into the second class, not the first. For completed integer ages, the result matches the 10–11 / 12–13 / 14–15 grouping above.
Why a missing value is not an age category
Suppose there were actually 32 pupils present but only 30 supplied a valid age. The repaired age groups still count 30 valid responses. The other two are missing age responses, not automatically “under 10”, “over 15” or part of “14–15”. A report could say “32 pupils approached; 30 valid age replies; two missing.” The group percentage denominator is 30 if the question is the distribution among valid replies.
If instead the question is “What share of the full class supplied a valid age?” the denominator changes: 30/32 × 100 = 93.75%. These are two different measures. A class should not substitute one for the other without changing its label.
Completed age versus exact age
“Age in completed years” means a pupil who has turned 12 but not yet 13 is recorded as 12, assuming the survey asks for completed years at a specified date. The underlying exact age is a continuous time quantity, but the questionnaire has grouped it into whole-year values. The decision affects what interval labels mean.
For exact ages, a label like “10–11” could be ambiguous if it is read as meaning only exact times from 10.000 up to 11.000. A clearly written interval 10 ≤ age < 12 includes ages such as 11.5 years and excludes age 12 exactly. The writer must define the boundary convention rather than assume everyone interprets the dash in the same way.
This matters with other measures too. Heights might be 140.0, 149.9 or 150.0 cm. An interval 140 ≤ height < 150 includes 149.9 cm but not 150.0 cm; the next interval 150 ≤ height < 160 includes exactly 150.0 cm. Nothing should fall between these consecutive bins if the units and precision are handled consistently.
How to audit a grouped table step by step
- Read the original measurement definition, unit and precision.
- Write every class boundary clearly before counting.
- For each boundary value, check which one and only one class contains it.
- Test values below the smallest boundary and above the largest; add an appropriate group if these are valid observations.
- Place each valid row into one group using the same rule.
- Sum all frequencies and check against the valid-record total.
- Retain a small check table or coding rule so someone else can reproduce the frequencies.
The boundary audit should precede the attractive chart. If a class histogram or bar chart looks convincing but the bins overlap, its neat presentation cannot make the underlying counts correct. A reader should be able to trace a single example pupil from their recorded age into exactly one class.
Progressive practice: four stages
Stage 1: one boundary value
Given the proposed groups 8–10 and 10–12, ask where age 10 belongs if both endpoints are inclusive. It fits both. Then ask for a repair: 8–9 and 10–12 for whole-number ages, or 8 ≤ age < 10 and 10 ≤ age < 13 if the intended upper bound is 13. The student must explain the revised coverage rather than merely cross out a number.
Stage 2: a counting check
A fictional dataset contains ages 9, 9, 10, 10, 10, 11, 12 and 12. Group them as 9–10 and 11–12 for completed whole years. The counts are 5 and 3, total 8. If the learner produces nine or ten entries, ask them to locate the duplicated raw observation.
Stage 3: independent grouping
Provide completed ages 11, 11, 12, 13, 13, 14, 14, 15, 15, 15. Students must invent two or three consecutive non-overlapping groups that cover all ten values, then show each frequency and total. There is more than one valid grouping. Assess its consistency, completeness and the clarity of the labels rather than demanding one arbitrary set of brackets.
Stage 4: transfer to height data
A fictional science table records heights 149.9 cm, 150.0 cm, 150.2 cm and 159.9 cm. Using 140 ≤ h < 150 and 150 ≤ h < 160, the frequencies are 1 and 3. A height of 150.0 cm belongs in the second group. Now ask students to explain why “140–150” and “150–160” without a boundary rule are insufficiently precise.
When should groups have equal widths?
For many histograms, equal-width class intervals make comparisons straightforward. A special uneven grouping can still be meaningful when it answers a specific question, but its graphical representation needs care: in a conventional histogram with unequal bin widths, area rather than height alone represents frequency. Beginners should generally start with a simple frequency table and equal-width bins unless a justified task calls for something else.
Do not choose groups after seeing the outcomes simply to make one category look unusually large. Decide the measurement question, keep the grouping transparent, and tell readers if the intervals changed between two surveys. An apparent trend can be a grouping artefact rather than a genuine change.
What parents and teachers can check
A parent does not need a statistical package to help. Ask: “Where would exactly 12 go?” “Can any value belong to two groups?” and “Does the total equal the number of pupils who actually answered?” These three checks often reveal the error. Parents can then offer a fresh dataset and ask the learner to construct new labels independently.
Teachers can ask students to keep a two-column trail: original value and assigned class. The trail need not reveal sensitive pupil identities; anonymous codes or synthetic data work well. Progress is visible when learners identify an overlap themselves, repair it, preserve valid missing counts separately and justify the resulting table without prompting.
Make the correction specific. If the frequency sum is too high because pupils were counted twice, do not merely tell the class to “be more careful with addition”. Show which shared endpoint made the categories non-exclusive. If the sum is too low, check for a gap in intervals or genuinely missing observations.
Limits: when multiple categories are intentional
Some questionnaires allow a pupil to tick several activities—football, chess and music, for example. In that kind of multi-response table, totals can exceed the number of respondents legitimately. The rule “frequencies must add to the number of pupils” applies when every valid observation is assigned to exactly one exclusive age interval. A responsible learner identifies the response design before applying a total check.
Age collection may also raise privacy considerations. For a classroom task, use the minimum necessary personal data, avoid posting identifiable pupil ages with names, and prefer simulated records for public projects. The statistical principle can be taught without distributing classmates’ personal details.
FAQs
Can 12 be in both 10–12 and 12–14?
It can under inclusive mathematical interpretations, which is exactly why those classes fail as exclusive age groups. Define non-overlapping intervals so 12 belongs to one class.
What is the easiest label for whole-number ages?
For the example, 10–11, 12–13 and 14–15 works clearly when the table says the values are ages in completed years and the integer endpoints are included.
Should I remove age 12 to fix the overlap?
No. Age 12 is a valid observation. Repair the class definitions, not the pupil data. Do not discard inconvenient values simply to force the totals to fit.
What if a pupil’s age is outside all groups?
Expand the set of groups or report a documented out-of-range category as appropriate to the question. Do not quietly omit a valid observation.
Do grouped frequencies show the exact average age?
Not necessarily. Grouping loses some detail. If original ages are available and collecting them was appropriate, the exact mean of those recorded values can be calculated from the raw data; estimates from grouped midpoints can differ. Do not present a grouped midpoint estimate as an exact raw-data mean.
A clear table is an act of good reasoning
For connected study, see how changed survey response options affect comparisons and how inequalities show whether endpoints are included. The eduKate Education Hub connects this skill across disciplines. The Clementi Mathematics tutorial reference models the habit of checking an answer independently.
The core lesson is simple but powerful: a trustworthy table begins with categories that tell each value exactly where it belongs. The final total is a check on that reasoning, not a substitute for it.