To calculate the mean from a frequency table, multiply each value by its frequency, add those products, then divide by the total frequency. In symbols: mean = Σ(fx) ÷ Σf. The catch that trips people up is grouped data: when the table gives class intervals instead of single values, you estimate using each interval's midpoint, so the result is an estimate, not an exact mean. This guide walks both cases with numbers you can check line by line.
Read the Frequency Table's Three Columns Before Touching a Calculator
Every frequency table you meet in a GCSE, A-level, or intro-statistics course follows one shape: a column of values (x), a column of frequencies (f), and optionally a working column (fx). The frequency tells you how many times a value appears in the raw data. If the table says a score of 4 has frequency 7, that means seven people scored 4 — not that 4 appears once with weight 7. Read the columns left to right and confirm what the top row labels mean before doing arithmetic.
Check whether frequencies sum to a sensible total. A survey of 40 students should produce frequencies adding to 40. If they don't, either the table is incomplete or you're looking at a two-way table instead. A quick sum in the margin catches most errors before they propagate. The Σf figure you get here becomes the denominator for every calculation that follows.
Distinguish population tables from sample tables only when you move on to variance. The mean formula is identical either way. What changes is the divisor in later steps — n versus n−1 — and that's a separate problem. For the mean alone, you divide by Σf regardless of whether the data represents a whole population or a sample drawn from one.
One more reading check: look at whether the values are discrete (whole numbers, shoe sizes, goals scored) or continuous (heights rounded to the nearest centimetre, times recorded to the nearest second). Continuous raw values have already been grouped before they hit the table. Spotting this early tells you whether to expect an exact answer or an estimate.
Multiply Each Value by Its Frequency to Build the fx Column
The fx column is the engine of the whole calculation. For each row, write x × f in a new column beside the frequency. A row with value 3 and frequency 8 contributes 24 to the running total. Do this row by row, keeping the table's original order so you can trace mistakes later. There's no shortcut that stays accurate for anything beyond three or four rows.
Mental arithmetic works for small numbers, but write the products down. A single slipped multiplication — 6 × 7 written as 41 instead of 42 — corrupts the final answer by a fraction that looks plausible and hides in the decimal places. Slower is faster here. If you're doing this on an exam paper, leave the fx column visible; examiners award method marks for it.
Sum the fx column carefully, then cross-check against the frequency total. If Σf is 40 and the fx sum comes out to 512, the mean is 512 ÷ 40 = 12.8. Confirm that the mean sits between the smallest and largest values in the table. If your value column runs from 2 to 20 and you get 12.8, that passes the smell test. A mean of 0.7 or 34 signals an arithmetic fault.
Double-check by working the sum twice, ideally in a different order — bottom-up if you went top-down. This catches transposition errors that a repeat in the same direction misses. The two-minute habit is standard practice among anyone marking these tables by hand.
Divide the fx Total by the Frequency Total for the Exact Mean
The mean is Σ(fx) ÷ Σf. Keep the numerator and denominator in full until the final division, then round as instructed. If a table shows 200 students and their Σ(fx) is 3,480, the mean is 17.4. Round to one decimal for reporting unless the question specifies otherwise. The unit matters too: if the values are test scores, the mean is a score; if they are hours of sleep, the mean is hours.
Know when to round. Continuous data often warrants one or two decimal places; discrete counts usually report clean numbers. A mean of 3.5 goals per match is odd to report as an integer. A mean of 12.8 correct answers out of 20 is fine as it stands. Match the rounding to whatever the exam or report specifies — the underlying calculation doesn't need extra precision beyond that.
Handle calculator entry in one pass. Type the sum of fx, press divide, type the frequency total, and hit equals. Splitting it into two calculator steps invites a transcription error. If you must take an intermediate figure, keep three or four decimal places before rounding the final answer.
If a table has a "total" row, use it. That row should already contain Σf and Σfx — cross-check your own sums against it. Where the totals disagree, trust neither until you've re-added by hand. Publishing a mean built on a mis-added total is the most common error that survives to the final output.
Estimate the Mean From Grouped Data Using Class Midpoints
When the table shows intervals like 10–19, 20–29, 30–39 rather than single values, the exact mean is unrecoverable. You work with midpoints instead. The midpoint of 10–19 is 14.5 (or 14 if the class is defined as 10 ≤ x < 20). Multiply each midpoint by its frequency, sum those products, and divide by the frequency total. The result is an estimate — standard practice, accepted in every exam board's mark scheme across the UK.
Understanding why midpoints work helps you spot errors. Inside each class, the raw values cluster around the centre, so the midpoint approximates the average of that class. If values within a class are evenly spread, the midpoint is exact. In practice they're rarely perfectly spread, so small errors creep in — but with reasonably sized classes the estimate lands close to the true mean.
Extract the midpoint with care. For a class written "20 to under 30" or "20 ≤ x < 30", take 25. For "20–30" that includes both endpoints, take 25 as well (the difference is negligible at these class widths). The trap is a mismatched class width — if one row spans 10 units and the next spans 15, average the endpoints, don't just eyeball the middle.
State clearly in your working that the answer is an estimate. A short note like "estimated mean = 27.3, using class midpoints" protects you in exams and prevents a reviewer confusing an estimate with an exact figure in a report.
Build the Full fx Working Column for a Grouped Table Step by Step
The method scales the same way for grouped tables, but the working column is where errors hide. A table of daily screen time with intervals and frequencies produces a larger numeric workspace. Write out each step under a heading so the audit trail stays visible.
1: Copy the frequency table and add two blank columns: "midpoint x" and "frequency × midpoint (fx)". 2: Fill the midpoint column by averaging each class's lower and upper bounds. 3: Multiply midpoint by frequency for each row; enter in the fx column. 4: Sum the frequency column to get Σf. 5: Sum the fx column to get Σ(fx). 6: Divide Σ(fx) by Σf and round to the requested precision.
If your class intervals contain a gap — say 10–19 then 21–29, with no 20 — the midpoint method still works, but flag it. Missing boundary values occasionally indicate the original data was rounded, and the gap is a rounding artefact rather than missing observations. State your assumption in one line, then carry on. In an exam, the mark scheme typically runs on midpoints regardless.
Check your result against the midpoint of the largest class. If your estimate lands outside the range of all midpoints, something has gone wrong. Suppose the midpoints run from 4.5 to 44.5 and your computed mean is 52 — recompute. The estimate must fall inside the range of available midpoints.
Work Through a Complete Example With Auditable Numbers
Here's a small grouped table from a hypothetical sample survey, using figures you can re-derive to confirm the method.
| Time (minutes) | Frequency (f) | Midpoint (x) | fx |
|---|---|---|---|
| 0–9 | 4 | 4.5 | 18 |
| 10–19 | 9 | 14.5 | 130.5 |
| 20–29 | 12 | 24.5 | 294 |
| 30–39 | 6 | 34.5 | 207 |
| 40–49 | 3 | 44.5 | 133.5 |
| Total | 34 | — | 783 |
The estimated mean is 783 ÷ 34 ≈ 23.03 minutes. Check that the answer sits between 4.5 (lowest midpoint) and 44.5 (highest midpoint) — it does. Re-add the frequencies: 4 + 9 + 12 + 6 + 3 = 34, matching the total row. Re-add the fx column bottom-up to cross-check: 133.5 + 207 + 294 + 130.5 + 18 = 783. Everything reconciles.
Now a discrete example on the same method. A frequency table of shoe sizes for 25 people has values 6, 7, 8, 9, 10 with frequencies 2, 5, 10, 6, 2. Σf = 25, and Σ(fx) = (6×2) + (7×5) + (8×10) + (9×6) + (10×2) = 12 + 35 + 80 + 54 + 20 = 201. The exact mean is 201 ÷ 25 = 8.04. No estimate here — this is the true mean for the sample.
Compare the two output formats. The grouped answer carries an "≈" and earns a method mark set; the discrete answer is exact and should carry an "=" sign. Confusing these two in an exam is a common way to lose an easy accuracy mark.
Spot the Five Errors That Send the Answer Wrong
Tables with a total row that doesn't match your Σf usually fail on one specific row — a frequency typed from the wrong column, or a reading of "15" when the original said "1.5". Track it by summing frequencies in pairs and locating which pair breaks the pattern. Rectify the row, rebuild the fx column, and discard any working built on the fault.
Getting the midpoint wrong for an uneven class is the second big fault. Classes of unequal width trip up anyone averaging "the middle number" without checking bounds. Always compute (upper + lower) ÷ 2 to avoid the trap. If the class is bounded unequally at the top — "60 and over" — you cannot calculate a midpoint; the mean from that table is undefined without an assumption about the upper bound, and any answer requires you to state that assumption plainly.
A third source of error is dividing by the number of rows instead of the frequency total. If the table has 6 rows and Σf = 60, the wrong divisor is 6 — producing an answer ten times too large. Cover this by writing "Σf = 60" on its own line before the division.
Rounding too early produces a plausible-looking answer that fails checks on a spreadsheet. If you round 783/34 to 23 and then need to use it in a later step, the accumulated error grows. Keep three or four decimals until the final line.
Finally, watch for a table where a frequency is zero. That's fine — a value that doesn't appear in the dataset. Keep it in the table, contribute zero to the fx column, and don't inflate the divisor. Deleting zero frequencies does nothing mathematically, but it hides the fact that the value existed.
Confirm the Answer Three Ways Before You Publish It
Run three checks in sequence. First, range check: the mean always lies between the smallest and largest values (or midpoints). Second, recomputation via a different order: sum the fx column top-down and bottom-up and compare. Third, frequency reconciliation: Σf must match the recorded sample size or total row. Any failure in one of these three stops the answer from going forwards.
For grouped tables, a fourth check works well: compare the estimated mean against the raw-data mean if you happen to have it. If the raw mean is 23.5 and your midpoint estimate is 23.03, the gap of about 0.5 is expected — midpoints compress within-class variation. If the raw mean is 30, something has gone wrong in the midpoint extraction.
In a spreadsheet, use =SUMPRODUCT(values, frequencies)/SUM(frequencies) to automate the row-by-row multiplication and summation. For grouped tables, use a helper column holding the midpoints, then the same SUMPRODUCT. The advantage is that you can click any cell and audit the formula. Reading a wrong answer back into a report becomes much harder.
When the result is being reused — in a chart, a report, or a downstream calculation — save the working table alongside it. Anyone auditing later will ask how the number was produced. A visible fx column answers that question in five seconds.
Answer Edge Cases: Zero Frequencies, Missing Upper Bounds, and Weighted Feels
While the arithmetic is straightforward, some tables resist the standard method. Here's a short diagnostic list.
Open-ended top class ("50+"): no midpoint possible without an assumption. Choose a sensible bound (e.g., 50–60) and disclose it. The answer becomes explicitly conditional. Uneven class widths: always calculate (lower + upper) ÷ 2 for each class — never assume a fixed width. Zero frequencies: keep the row, contribute zero to fx, don't delete. Frequencies given as decimals: the method still works; Σf may not be an integer and the mean is still Σ(fx) ÷ Σf. Raw data vs frequency table disagreement: trust the raw data if both are provided. If only the table is provided, the answer is derived entirely from the table.
A subtle edge case is a table with cumulative frequencies ("less than 20: 8, less than 30: 17"). Convert back to individual frequencies first by differencing adjacent rows, then run the standard mean calculation. Going straight into multiplication with cumulative figures is the single most common error on these tables.
Stating an assumption protects you. A sentence like "the estimated mean assumes values are evenly spread within each class" is standard practice in statistics reporting. It signals you know the limits of the midpoints approach, without undermining the calculation.
FAQ
Can you calculate the exact mean from a grouped frequency table?
No. Once continuous data is grouped into classes, the raw values are lost and only an estimate is possible. The estimate uses class midpoints as proxies for each value, and it's typically within a small band of the true mean — but it's not the mean. State the answer with "≈" and note the assumption of even spread within each class.
What's the difference between a frequency table and a grouped frequency table?
A frequency table lists individual values beside how often each occurs (e.g., 7 appears 5 times). A grouped frequency table collapses ranges into classes (e.g., 20–29 has frequency 12), usually because the raw data is continuous or too spread out to list row by row. The mean calculation differs only in whether you multiply frequencies by values or by class midpoints.
How do I handle a frequency table with a "50 or more" class at the top?
The midpoint can't be calculated without choosing an upper bound, so any mean is conditional on that assumption. Common practice in exam papers is to specify the recount — for example, treat "50 or more" as 50–59 — but in real reports you should state the assumption explicitly. Without it, the mean of that table is undefined.
Why is my answer wrong when I divide by the number of rows?
Dividing by the number of rows instead of Σf shrinks the divisor and inflates the mean, sometimes by an order of magnitude. A table with 6 rows and Σf = 60 needs division by 60, not 6. Write "Σf =" on its own line every time to lock the divisor in before calculating.
Can I use a calculator's statistics mode to speed this up?
Yes — on most scientific calculators, enter each value (or midpoint) with its frequency and let the calculator report the mean directly. The check remains identical: sum of fx over sum of f. On a Casio or Texas Instruments model, this typically saves 30–60 seconds per table. Still cross-check by hand for one row to confirm your data entry is correct.
What does the answer mean if my frequencies come from a sample rather than the whole population?
For the mean, the formula is the same — Σ(fx) ÷ Σf. The distinction only matters when you compute variance, where the sample formula divides by (n − 1) rather than n. The mean is the same arithmetic either way; only the interpretation of the result changes (sample average versus population average).
Can I derive the mean from a two-way table?
Yes, but only by collapsing one dimension first. Sum the frequencies across the other variable and build a one-way frequency table of values against totals, then run the standard method. A two-way table showing hours studied by year group, for example, becomes a single frequency table of hours once you add frequencies across years — but only if the row and column totals match the raw data throughout.