Can absolute frequencies be used for numerical data?

Yes. They can count exact values or observations within defined intervals.

Are raw absolute frequencies integers?

Ordinary unweighted counts are nonnegative integers. Weighted estimates are different quantities and should be labeled accordingly.

Does a larger count mean a higher proportion?

No. The total number of observations may also have changed.

AI Glossary / Statistics and probability / Beginner

Absolute frequency

Definition

Absolute frequency is the number of times a specified value, category, or event occurs in a defined dataset. It is a count, such as eight requests in a particular category. Unlike relative frequency, it is not divided by the total number of observations. A meaningful absolute frequency requires a clear counting unit, observation period, inclusion rule, and treatment of missing or repeated records.

What it means

If a dataset contains five observations labeled delivery, the absolute frequency of delivery is five. The count describes what was observed within that dataset; it does not by itself show how common delivery is relative to other categories or in a wider population.
The word absolute distinguishes the raw count from a proportion or percentage. It does not mean perfectly accurate, free from uncertainty, or independent of data-collection choices. Counting requests, customers, and sessions can produce different frequencies from the same underlying activity.

How it works

Define the unit to be counted and the population or dataset covered. Decide whether repeated events from the same person are separate observations and how duplicate records will be identified. Specify the time period and any exclusions.
Assign observations to values or categories using consistent rules. For grouped numerical data, define interval boundaries so observations are not inadvertently counted twice or omitted. For categorical data, standardize equivalent labels before counting when the analysis requires them to represent the same category.
Count occurrences in each category. In a mutually exclusive and exhaustive classification, category counts sum to the total number of observations. In a multiple-response question, one person can contribute to several category counts, so their sum can exceed the number of people.
Report counts with their context. Relative frequency can be calculated by dividing a count by an appropriate total, but the denominator must be explicit. Missing values can be shown as a separate category or excluded for a specified calculation; those choices answer different questions.

Three practical examples

A support team counts requests in each issue category during a month to plan workload.
A machine-learning analyst counts the examples assigned to each class before training a classifier.
A survey team reports how many respondents selected each option, noting whether respondents could choose more than one.

Think of it this way

Think of tally marks on a sheet. Each mark represents one occurrence under a chosen rule. The tally tells you how many marks there are, but understanding what they mean requires knowing what earned a mark, which observations were eligible, and whether the same unit could contribute more than once.

Worked example

Consider an invented dataset of 20 support requests. Eight are labeled billing, five access, four delivery, and three have no recorded category. The absolute frequencies are therefore 8, 5, 4, and 3, and their sum is 20.
Billing accounts for 8 / 20 = 40% of all requests. If the question instead concerns the distribution among requests with a recorded category, the denominator is 20 − 3 = 17. Billing then accounts for 8 / 17, approximately 47.06%. The absolute billing frequency remains eight in both calculations.
In the next month, suppose there are 40 requests and ten are labeled billing. The absolute billing count rises by two, or 2 / 8 = 25% relative to the earlier count. Yet billing's share of all requests falls from 40% to 10 / 40 = 25%, a decrease of 15 percentage points.
These statements are consistent: total request volume increased enough that the category's share fell despite a higher count. The example assumes the same counting and labeling rules across months. It does not establish that customers became more or less likely to experience billing problems without information about customers and exposure.

Illustrative case study

In this fictional case, Pine Helpdesk sees a rise in the number of requests labeled access and considers adding staff. Before deciding, analysts inspect the counting rule and discover that one system records each customer reply as a new row, while another records one row per complete request.
The team reconciles the units, separates duplicate records from genuine repeat requests, and reports both workload counts and the share of categorized requests. It keeps missing categories visible so an apparent trend is not driven by a change in labeling completeness.
The revised report provides a clearer basis for staffing discussions while acknowledging that request counts alone do not measure customer satisfaction or underlying failure rates.

Common misconceptions and mistakes

Absolute frequency is not a percentage, probability, or rate per unit time or exposure. A count of ten events needs a denominator before it can support many comparisons.
Another mistake is assuming category counts always sum to the number of people. Multiple-response data, overlapping categories, and repeated events can make that false. Conversely, missing or excluded records can make totals smaller than expected.
A recorded zero also differs from missing information. Zero means no occurrences were counted under the stated observation process; it does not prove that the event is impossible or absent from an incompletely observed population.

Business use

Counts support planning for volume, inventory, staffing, and data preparation. They are also useful for checking class imbalance and identifying categories with too few observations for reliable analysis.
Use consistent definitions across time and sources. Pair counts with appropriate proportions or exposure-based rates when comparing groups of different sizes. Preserve the underlying counting rules so readers can understand changes and reproduce the figures.

Check your understanding

What is billing's absolute frequency in the first example month?

Answer: Eight.

Why are 40% and 47.06% both possible?

Answer: They use different denominators: all requests and categorized requests.

Can a count rise while its share falls?

Answer: Yes, as the second-month example demonstrates.

Related terms

Relative frequency
Frequency distribution
Proportion
Missing data
cumulative frequency
count
denominator
class imbalance

Learn next

Learn next how frequency tables distinguish counts from percentages. Study multiple-response data and exposure-based rates, then examine how sampling and data quality affect conclusions drawn from observed counts.

Data Quality

Sources

Australian Bureau of Statistics, Describing Frequencies: https://www.abs.gov.au/statistics/understanding-statistics/statistical-terms-and-concepts/describing-frequencies Australian Bureau of Statistics, Frequency Distribution: https://www.abs.gov.au/statistics/understanding-statistics/statistical-terms-and-concepts/frequency-distribution Australian Bureau of Statistics, Statistical Terms and Concepts Glossary: https://www.abs.gov.au/statistics/understanding-statistics/statistical-terms-and-concepts/statistical-terms-and-concepts-glossary

Previous: Absolute error loss Next: Absolute position