Absolute and Relative Frequency

Absolute frequency counts how often a value or category occurs, while relative frequency divides that count by the total number of observations. In brief: if 6 of 20 observations belong to a category, its absolute frequency is 6 and its relative frequency is 6/20 = 0.3, or 30%. The two measures describe the same category on different scales.

Frequency tables organise raw data so distributions can be compared and interpreted. Counts show how many observations occurred, while proportions account for the size of the dataset. This guide explains table construction, percentage and cumulative forms, grouped intervals and the limits of what observed frequencies can tell you.

What is absolute frequency?

The absolute frequency of a value or category is its number of occurrences in a dataset. If the value 3 appears five times in a list, its absolute frequency is 5. A frequency is a count, so it is a nonnegative integer for ordinary unweighted data.

Categories may describe qualities, such as colours, or numerical values, such as numbers of books borrowed. Each observation is assigned to a category or value according to the table’s definition. The category labels should be clear enough that the same observation would be classified consistently by different readers.

If the categories are mutually exclusive and cover all observations, their absolute frequencies sum to the sample size n. This total is a useful check for missed or double-counted entries. A zero frequency means the category was included in the table but did not occur in the observed data.

What is relative frequency?

Relative frequency is the category count divided by the total: f/n, where n is positive. It can be written as a fraction, decimal or percentage. A count of 8 in a dataset of 40 gives 8/40 = 0.2 = 20%, allowing comparison with datasets of different sizes.

For a complete set of non-overlapping categories, relative frequencies sum to 1, or 100% in percentage form. Rounding may make the displayed total slightly different, such as 99.9% or 100.1%. The exact unrounded ratios still sum to 1 when all observations are accounted for.

Relative frequency is dimensionless because the count in the numerator and the total count in the denominator refer to the same kind of observation. It describes a share of the dataset, not an additional number of items. To recover a category count from an exact relative frequency, multiply by n.

How do you build a frequency table from raw data?

Choose categories or list the distinct observed numerical values. Then count how many observations fall into each row, using tally marks or another systematic method. Add the counts to verify the total before calculating proportions, because an incorrect total affects every relative frequency in the table.

Suppose 20 recorded responses are classified as A, B, C or D, with counts 6, 8, 4 and 2. Divide each by 20 to obtain relative frequencies 0.30, 0.40, 0.20 and 0.10. The counts total 20 and the proportions total 1.

The table should identify what was observed and the size of the dataset. Category order may be alphabetical, logical or based on the measured variable; use an order that helps interpretation. Do not imply a numerical ranking for purely nominal categories unless the labels themselves have an established order.

1. Define mutually exclusive categories covering the data.
2. Count the observations in each category.
3. Verify that the counts sum to the sample size.
4. Divide each count by the total.
5. Label fractions, decimals or percentages clearly.

What is cumulative frequency?

Cumulative frequency is a running total through ordered categories or values. If frequencies for values 0, 1, 2 and 3 are 2, 5, 4 and 1, their cumulative frequencies are 2, 7, 11 and 12. The entry 11 means eleven observations have values at most 2.

Cumulative relative frequency divides each running total by the overall sample size. In the example, the share at most 2 is 11/12, approximately 91.67%. The final cumulative frequency must equal n, and the final cumulative relative frequency must equal 1 before rounding.

Cumulative measures require a meaningful order. They are useful for numerical values or ordered categories, but an alphabetical running total of favourite colours usually has little substantive meaning. The calculation is possible, yet its interpretation depends on what moving through the category order represents.

How do grouped frequency intervals work?

When many numerical values occur, group them into intervals instead of listing every distinct value. For example, intervals 0 ≤ x < 10, 10 ≤ x < 20 and 20 ≤ x < 30 place each value in exactly one class. A value of 10 belongs to the second interval, not both the first and second.

State interval endpoints clearly and avoid gaps or overlaps. If the final maximum equals a boundary, define the last interval to include it or extend the range appropriately. The categories must cover the actual data, and their boundary convention should be visible in the table labels.

Grouping loses information about exact values within each interval. It can reveal an overall distribution more clearly, but a grouped table cannot usually reconstruct the original dataset. If class widths differ, raw frequencies also cannot be compared as histogram heights without accounting for width through an appropriate density scale.

Why are relative frequencies useful for comparing groups?

Absolute counts can be misleading when group sizes differ. If 12 of 20 observations in one group meet a condition and 30 of 100 in another do, the second group has more such observations but a smaller share. The relative frequencies are 60% and 30%, respectively.

To combine groups, add their counts and totals before dividing. The combined share is (12 + 30)/(20 + 100) = 42/120 = 35%. Averaging 60% and 30% directly would give 45%, incorrectly giving the two unequal groups the same weight.

Comparability also requires consistent category definitions and observation procedures. A difference in relative frequency may reflect a real distribution difference, a changed definition or sampling variation. The table describes the observed data; explaining why the groups differ requires information beyond the arithmetic alone.

How is relative frequency related to probability?

Relative frequency records an observed proportion in a finite dataset. Probability describes a modelled chance for an event under stated conditions. Repeated observations can be used to estimate probability, but an observed relative frequency is not automatically the exact probability of a future event.

If a repeatable experiment has an event probability p and suitable independence conditions, larger samples tend to give relative frequencies closer to p in a probabilistic sense. Small samples can fluctuate substantially. A category observed zero times therefore need not be impossible; it may simply be uncommon in the limited data.

This connection provides the background for experimental probability and statistical inference. Frequency tables first summarise what happened, while a probability model or inferential method asks what those observations suggest beyond the dataset. Keeping description and modelling distinct prevents a sample proportion from being presented as certainty.

What mistakes should you avoid with frequency calculations?

Use the correct denominator. A relative frequency for the whole dataset divides by the full sample size, while a conditional share within a subgroup divides by that subgroup’s total. Both can be valid, but they answer different questions and must be labelled accordingly.

Do not add relative frequencies from overlapping categories as though they partitioned the data. One observation may belong to several categories, so their sum can exceed 100% without an arithmetic error. State whether the table permits multiple responses or assigns each observation to exactly one row.

Finally, preserve counts alongside rounded percentages whenever possible. Counts allow readers to reconstruct exact proportions and assess the size of the dataset. Check totals, category coverage and rounding, then describe patterns in terms of the observed data rather than claiming a precision or generality that the table does not establish.