The relative frequency calculator above turns a raw list into the table that statistics courses build by hand in week one. It finds every distinct value, counts how often each appears, divides each count by the total, shows that proportion as a percentage, and runs two cumulative columns down the table so you can read off how much of the data sits at or below any point. Text labels are accepted as well as numbers, because relative frequency is taught as often on colours, grades and survey answers as on measurements.
Arb Digital publishes this alongside the rest of the statistics set because the frequency table is the step most people skip. Summary statistics compress a data set into one or two numbers, which is useful and lossy. The table keeps the shape intact: where the mass sits, whether there are two peaks, whether one category carries most of the total.
What This Relative Frequency Calculator Does
Frequency is a count. Relative frequency is that count divided by the number of observations, so it always lands between 0 and 1 and the whole column adds to 1 by construction. Multiply by 100 for the percentage form. The calculator shows all three, because each suits a different sentence: counts for "how many", proportions for arithmetic, percentages for writing up.
The two cumulative columns are the part free tools most often leave out. Cumulative count is a running total of the counts; cumulative relative frequency is a running total of the proportions, finishing at exactly 1. That final column answers a different question — not "how many were 8" but "what share of the data was 8 or less". The bar breakdown under the table scales every category against the largest, so the shape is visible before you read a figure.
How to Use It
- Paste your data into the box. Commas, spaces, tabs, semicolons and line breaks all work in any combination, so a spreadsheet column needs no cleaning up first.
- Check the n figure against the number of observations you expect. A mis-paste or a duplicated block shows up here immediately, which is why n is on screen.
- Choose your decimal precision. Two suits presentation; four is worth switching to when several categories are small, because at two decimals different proportions can print as the same number.
- Read the table. Each row gives the count, proportion, percentage, running count and running proportion.
- Use the row order control to switch between value order, which the cumulative columns need, and count order, which puts dominant categories first.
The Formula: How Relative Frequency Is Calculated
For a category with count f in a data set of n observations, the relative frequency is f ÷ n and the percentage is (f ÷ n) × 100. There is no sample-versus-population choice here and no corrected divisor of the kind the variance calculator has to offer: you are describing the data you have, not estimating a parameter, so the denominator is simply the number of observations.
The cumulative relative frequency at any row is the sum of the relative frequencies of that row and every row above it. Equivalently it is the cumulative count divided by n — a useful identity, because the whole cumulative column can then be rebuilt from the counts alone with no rounding involved. The calculator computes it from the exact counts for that reason.
Two constraints follow. Every relative frequency lies between 0 and 1, so a value outside that range means an arithmetic error, and the column sums to exactly 1 because the counts sum to n. Those two checks catch most mistakes made by hand.
A Worked Example You Can Check Yourself
The calculator loads with twenty observations: 3, 5, 5, 7, 7, 7, 7, 8, 8, 8, 9, 9, 10, 10, 10, 10, 10, 12, 12, 15. Counting them off gives eight distinct values — 3 once, 5 twice, 7 four times, 8 three times, 9 twice, 10 five times, 12 twice and 15 once. Those counts add to 1 + 2 + 4 + 3 + 2 + 5 + 2 + 1 = 20, which matches n, so nothing has been missed or double-counted.
Divide each count by 20. One twentieth is 0.05, so the proportions are 0.05, 0.10, 0.20, 0.15, 0.10, 0.25, 0.10 and 0.05 — that is 5%, 10%, 20%, 15%, 10%, 25%, 10% and 5%. Adding them in order gives 0.05, 0.15, 0.35, 0.50, 0.60, 0.85, 0.95 and finally 1.00, the cumulative relative frequency column ending where it must. The cumulative counts run 1, 3, 7, 10, 12, 17, 19, 20.
The headline result is the value 10, which appears five times and so accounts for 25% of the data. That makes it the mode. Only 7 comes close, at four occurrences — but 25% against 20% is one observation's difference, worth remembering before calling 10 typical.
Reading the Median Off the Cumulative Column
The cumulative column exists so you can answer position questions without sorting the raw list again. Look for where it crosses 0.50: that is the middle of the data. In the example it reaches exactly 0.50 at the value 8, which tells you the tenth observation in sorted order is an 8 and that half the data is 8 or less.
An exact 0.50 is the one case that needs care. With n = 20 the median is the average of the tenth and eleventh values — 8 and 9 — so the median is 8.5, a number that does not appear in the data at all. When the column hits 0.50 exactly the median sits on the boundary between that row and the next; when it crosses 0.50 partway through a row, the median is that row's value outright. Confusing the two is the commonest error people make reading a median off a cumulative table.
The same reading works at any threshold: where the column first passes 0.25 you have the first quartile, and where it passes 0.75 the third. That is the idea behind the percentile calculator, which applies an explicit interpolation convention rather than leaving you to eyeball the crossing point. This calculator builds the distribution; if you want the same data reduced to single centre values, use the mean median mode calculator instead.
Relative Frequency Is an Estimate, Not the Probability
This is the point most pages skip, and the one that matters. A relative frequency is an empirical quantity: what your sample did. A probability is a property of the process that generated it. They are different objects, and the first is used as an estimate of the second.
The law of large numbers licenses that use: as the number of observations grows, the relative frequency of an outcome converges on its true probability. Toss a fair coin twenty times and you may well see twelve heads, a relative frequency of 0.60; toss it twenty thousand times and the proportion sits very close to 0.50. Convergence is guaranteed in the limit, and says nothing reassuring about twenty observations.
So treat the 0.25 in the worked example with suspicion. It rests on five observations; one more or one fewer moves it to 0.30 or 0.20. A rough guide to the uncertainty in a proportion is the square root of p(1 − p) ÷ n, here the square root of 0.25 × 0.75 ÷ 20, about 0.097 — nearly ten percentage points of standard error on a figure quoted as 25%. The honest reading is that 10 is the most common value in this sample, not that a quarter of the population equals 10. This tool counts what happened; if you want to work forward from stated probabilities rather than observed counts, use the probability calculator instead.
Why the Column Sometimes Sums to 0.999
The relative frequencies always sum to 1 exactly. The displayed column sometimes does not, and that surprises people who assume an error has crept in. It has not.
Rounding is the whole explanation. Take three categories with counts of 1 out of 3. Each true proportion is 0.3333…, and each displays at three decimals as 0.333, so the column sums to 0.999. Neither that nor the 1.001 you get when several values round upward is a miscalculation: the underlying thirds still add to exactly 1, and only the printed version falls short. The results grid shows what the displayed proportions add up to, so the discrepancy is visible rather than something you notice later and worry about.
Two consequences. Never adjust one row to make the column tidy, because that row is then wrong by the amount you moved it; if a published table must total 100%, footnote that components may not sum due to rounding, as official statistics do. And compute anything downstream from the counts rather than the rounded proportions, so the error does not propagate. More decimals shrink the gap without eliminating it, which is the point: it is a display artefact and it lives in the display.
Grouped Data, Rare Categories and Where the Table Stops Helping
A frequency table works best when the number of distinct values is small relative to n. Twenty observations across eight values, as in the example, is comfortable. Twenty across nineteen is not: nearly every row shows a count of 1 and a relative frequency of 0.05, and the table has told you almost nothing.
Continuous measurements almost always land in that second case, because two measured heights or durations rarely repeat exactly. The fix is to group values into intervals — 0 to 10, 10 to 20 and so on — and count how many fall in each. That gives a readable histogram, at the cost of the bin width, which genuinely changes the picture: wide bins smooth away real structure, narrow bins invent it out of noise. This calculator counts exact repeated values rather than choosing bins for you, so for continuous data you would bin first and paste the labels in as text.
The other failure mode is a long tail of rare categories. When forty categories each hold one or two observations, keep the top handful and collapse the rest into an "other" row — but say so, since a table whose largest row is "other" has hidden its own subject. For counting words in a block of text rather than values in a list, the word frequency counter does the same counting job with tokenisation rules built for prose. The NIST/SEMATECH e-Handbook chapter on exploratory data analysis covers frequency tables, histograms and bin selection, and the Penn State Eberly College of Science statistics courses work through the same material with exercises.
Arb Digital builds analytics setups where the shape of the data stays visible, so decisions do not rest on one summary number.
Browse All Free Tools Talk To Arb DigitalCommon Mistakes to Avoid
- Treating a small-sample proportion as fixed truth — 25% from twenty observations moves five points if one data point changes.
- Adjusting a row so the percentages total exactly 100 — a rounded column summing to 99.9 is correct, and editing a figure to hide the gap makes it wrong.
- Reading a median off a cumulative column that hits 0.50 exactly — that marks a boundary between two values. With an even n the median is the average of the two.
- Building a table where almost every count is 1 — for continuous measurements, group into intervals first or the table carries no information.
- Reading the cumulative column after sorting by count — the running total only means "at or below" in value order.
Related Free Tools From Arb Digital
Reduce the same data to its centre with the mean median mode calculator, locate a single value with the percentile calculator, or measure spread with the standard deviation calculator. The free online tools hub lists the full statistics set.
Frequently Asked Questions
Relative frequency is the count of a value divided by the total number of observations. It always falls between 0 and 1, and the column always sums to 1. Multiplying by 100 gives the same quantity as a percentage.
Yes. Labels such as Red, Blue and Green are counted exactly like numbers, since relative frequency is most often taught on categorical data. Rows sort numerically when every entry is a number and alphabetically otherwise.
Rounding. Three equal categories each show 33.3%, totalling 99.9% even though the true proportions sum to exactly 1. It is a display artefact, not an error, and more decimals narrow the gap without removing it.
It gives the share of the data at or below each value, so positions can be read straight off the table. Where it crosses 0.50 is the middle of the data, where it crosses 0.25 the first quartile, and it always ends at 1.
No. Relative frequency is what your sample did; probability is a property of the process behind it. The law of large numbers says the two converge as the number of observations grows, which is why a proportion from a small sample carries real uncertainty.
Group them into intervals first and count how many fall in each, then paste the interval labels in. A table with one row per distinct value carries no information when almost every count is 1.
Neither applies. Relative frequency divides by the number of observations entered, because it describes that data set rather than estimating a parameter, so there is no corrected divisor of the kind variance uses.
This calculator counts and divides the data you paste in. Whether that data is a fair sample of anything wider is a separate question the tool cannot answer for you.
