Read frequency as a count of observations
To find the median from a frequency table, add the frequencies, locate the middle observation or observations, and use cumulative frequency to identify their values. A frequency is a count, not another measurement to average. If the value 4 has frequency 3, the underlying list contains three copies of 4. The median describes the center of that expanded ordered list, even when you never write the whole list out.
Begin by checking what the rows represent. This method gives an exact median when each row names one exact numerical value and its nonnegative whole-number count. A row labeled 10–19 represents an interval instead, so it does not reveal every observation inside that interval. Keep those two table types separate. The word frequency alone does not establish whether the data retain enough detail for an exact numerical median.
Sort the values and validate the table
Arrange the distinct values from smallest to largest before calculating running totals. Frequencies stay attached to their values during sorting. If a value appears in several rows, combine its counts or keep the equal-value rows adjacent. Sorting counts by size would change the question: the most frequent row can identify a mode, but its position does not establish the median. Negative values and decimals are allowed as observations when their order is clear.
Check every frequency before adding. Counts such as 0, 2 and 7 are valid; negative counts, fractional counts and blanks needing interpretation are not a complete exact-frequency record. A zero-frequency row contributes no observations and cannot hold a median position. If all frequencies are zero, there is no data set and therefore no median. Percentages or sampling weights require further information before they can be treated as literal observation counts.
Build the cumulative-frequency column
Consider an original example recording the number of support tickets handled in each of ten short sessions. The values are 1, 2, 4 and 7, with frequencies 2, 3, 4 and 1. The total is n = 2 + 3 + 4 + 1 = 10. Running totals are 2, 5, 9 and 10. Each cumulative total tells you the final rank occupied by that value in the sorted list.
Read each row as a block of positions. Value 1 occupies ranks 1–2, value 2 occupies ranks 3–5, value 4 occupies ranks 6–9, and value 7 occupies rank 10. The final cumulative count must equal n. That equality is a useful transcription check, but also inspect the individual rows: a wrong count carried consistently into the total can still satisfy the final equality while describing the wrong observations.
| Value | Frequency | Cumulative count / ranks |
|---|---|---|
| 1 | 2 | 2 / 1–2 |
| 2 | 3 | 5 / 3–5 |
| 4 | 4 | 9 / 6–9 |
| 7 | 1 | 10 / 10 |
The even-count median uses ranks 5 and 6, so the answer is (2 + 4) / 2 = 3.
For an odd count, locate one middle rank
When n is odd, the median rank is (n + 1) / 2 using positions that start at one. For a separate nine-session example, keep the same values but use frequencies 2, 3, 3 and 1. The cumulative counts become 2, 5, 8 and 9. The middle rank is (9 + 1) / 2 = 5, which falls in the value-2 block. The median is therefore 2 tickets per session.
At an exact cumulative boundary, keep the row whose cumulative count reaches that rank. Rank 5 belongs to the row ending at 5, not the next row starting at 6. This is why the lookup rule is to find the first cumulative frequency greater than or equal to the target rank. Do not use a strictly greater comparison. That small distinction matters whenever the central position lands precisely at the end of a frequency block.
For an even count, find both central values
Return to the ten-session example. An even count has central ranks n / 2 and n / 2 + 1, here 5 and 6. Rank 5 has value 2, while rank 6 has value 4. Average the two values: median = (2 + 4) / 2 = 3 tickets per session. The answer need not be an observed value. It is the midpoint between the two central observations under the usual numerical-median convention.
Do not average the central ranks themselves and report 5.5 as the measurement. That number describes a position between observations, not the ticket count. Equally, do not stop after locating the first row whose cumulative count reaches n / 2. Sometimes both middle ranks fall in the same row, yielding that row's value; sometimes they straddle a boundary. Looking up both ranks separately handles both cases without an extra memorized exception.
Verify the result with the raw-list calculator
For a small table, expand the ten-session record as 1, 1, 2, 2, 2, 4, 4, 4, 4, 7. Paste those observations into the Median and Mode Calculator. It sorts the values and reports the two middle positions, allowing you to compare its median of 3 with the cumulative-count method. It also reports the mode as 4, a useful reminder that the most frequent value and the central value answer different questions.
The current calculator accepts a raw list of 1–1,000 observations, not frequency rows. Entering 1, 2, 4, 7 once each would discard the counts and calculate a different data set. Entering the frequency column alone would instead find the median count. For totals above the list limit, retain the cumulative-position calculation or use an appropriate frequency-aware spreadsheet workflow rather than pretending the tool accepts a compressed table directly.
Recognize information that an interval table has lost
Suppose a row says that five observations lie between 10 and 19. Those observations might cluster near 10, near 19 or anywhere between. A cumulative count can identify the class containing the central observation, but cannot reveal its exact value. Replacing each class by its midpoint creates a new approximate data set. It does not recover the original median, and the midpoint of the median class is not automatically correct.
Similarly, rounded percentages may fail to identify an exact integer n or may hide a central boundary. Ask for the original counts when a precise result matters. A grouped-data median formula uses an interpolation assumption inside the selected class and should be labeled an estimate. For ordered categories without numerical spacing, locate a middle category where appropriate but do not average category labels as though their codes were measured quantities.
Document the calculation so another reader can check it
A concise median record includes the ordered values, original frequencies, cumulative frequencies, n, central ranks and final numerical values. For the ten-session example, write n = 10, ranks 5 and 6, values 2 and 4, median 3. State the unit alongside the answer. This record is more useful than an unexplained single number because another person can identify whether a disagreement came from counts, ordering or the even-count averaging step.
Use this educational method for exact numerical frequency records, such as repeated scores or item counts. Preserve legitimate extreme observations: the median's resistance to their size is not a reason to delete them. Compare the result with the mean or mode only when that comparison helps describe the data. If the source is incomplete or interval-grouped, report what is known and what is estimated instead of manufacturing precision from a tidy table.