Relative Frequencies — I can compute relative frequencies...

SECTION 1

Historical Context & Motivation

Long before spreadsheets and computers, people needed ways to organize and compare information about groups. Merchants tracked which products sold best in different regions, doctors recorded the outcomes of treatments for different populations, and governments collected census data about their citizens. The challenge was always the same: when you have raw counts of things, how do you make fair comparisons between groups of different sizes? A city with 500 flu cases sounds worse than a town with 50 cases—but what if the city has a million residents and the town only has 200?

1662

John Graunt's Life Tables

English statistician John Graunt published Natural and Political Observations on the Bills of Mortality, one of the first works to convert raw death counts into proportions, comparing causes of death relative to total deaths.

1785

Condorcet & Contingency Tables

French mathematician the Marquis de Condorcet used cross-classified tables to analyze voting patterns, organizing data by two categorical variables simultaneously—an early form of the two-way table.

1900

Karl Pearson's Chi-Square Test

Karl Pearson formalized methods for analyzing two-way tables, developing the chi-square test that relies on comparing observed relative frequencies to expected relative frequencies.

Today

Data Science & Machine Learning

Relative frequencies from two-way tables are foundational in modern data analysis, from medical research to social media analytics. Every time an app shows you a percentage breakdown, relative frequencies are at work.

The central question that relative frequencies answer is straightforward: what proportion of a total does each category represent? When we combine this idea with a two-way table—which organizes data by two categorical variables at once—we gain a powerful tool for spotting relationships and making meaningful comparisons in real-world data.

SECTION 2

Core Principles & Definitions

Before diving into calculations, let's nail down the vocabulary. A two-way table (also called a contingency table) displays data classified by two categorical variables. The rows represent one variable and the columns represent the other. Each cell shows the count (frequency) for that combination of categories. The totals along the right edge and bottom edge are called marginal frequencies, and the number in the bottom-right corner is the grand total.

Joint Relative Frequency

The ratio of a cell's count to the grand total. It tells you the proportion of all observations that fall into that specific combination of categories.

Marginal Relative Frequency

The ratio of a row total or column total to the grand total. It describes the overall distribution of one variable, ignoring the other variable entirely.

Conditional Relative Frequency

The ratio of a cell's count to its row total or column total. It answers "given that we're only looking at one category, what proportion falls into the other category?"

Relative vs. Absolute Frequency

An absolute frequency is the raw count. A relative frequency is that count divided by an appropriate total, expressed as a decimal, fraction, or percentage. Relative frequencies always sum to 1 (or 100%).

✦ KEY TAKEAWAY

Think of relative frequency like converting currency. If someone tells you a shirt costs 2,000 yen, that number alone doesn't help you decide if it's cheap or expensive—you need to convert it relative to something you understand (like dollars). Similarly, a raw count of 45 students doesn't mean much until you express it relative to the total. Relative frequency is your "conversion rate" from raw counts to meaningful proportions.

SECTION 3

Visual Explanation

Let's look at a concrete example. Suppose a school surveys 200 students about their preferred study method (alone or in a group) and whether they are in STEM or Humanities classes. The diagram below shows how a two-way table organizes this data, along with all three types of relative frequency.

This two-way table shows raw counts for 200 students classified by study preference and subject area. Joint relative frequencies divide each cell by the grand total (200). Marginal relative frequencies divide row or column totals by 200. Conditional relative frequencies divide a cell by its row or column total.

Notice how the same cell value of 70 produces different relative frequencies depending on which total you divide by. Dividing 70 by the grand total of 200 gives a joint relative frequency of 0.35, meaning 35% of all surveyed students are STEM students who study alone. Dividing 70 by the row total of 120 gives a conditional relative frequency of about 0.583, meaning roughly 58.3% of students who prefer studying alone are in STEM. The denominator you choose changes the story the number tells.

SECTION 4

Mathematical Framework

Each type of relative frequency follows a simple formula. The key is choosing the correct denominator. Let's define our notation: each cell count is represented by f, row totals and column totals are marginal frequencies, and N represents the grand total of all observations.

JOINT RELATIVE FREQUENCY

Joint RF = f(cell) ÷ N

Where f(cell) is the count in one specific cell and N is the grand total. This tells you the proportion of all observations that belong to a particular combination of categories.

MARGINAL RELATIVE FREQUENCY

Marginal RF = f(row or column total) ÷ N

Divide a row total or column total by N. This describes the overall distribution of one variable while ignoring the other. All marginal relative frequencies for one variable sum to 1.

CONDITIONAL RELATIVE FREQUENCY

Conditional RF = f(cell) ÷ f(row total) or f(cell) ÷ f(column total)

Divide a cell count by the total of the row (or column) it belongs to. This answers a "given that" question: given that a person is in a specific category of one variable, what proportion falls into a specific category of the other variable?

⚠️ Which Denominator?

The most common mistake students make is using the wrong denominator. Remember: joint → grand total, marginal → grand total, conditional → row or column total. If the question says "of all students," use the grand total. If it says "of students who…" or "given that…," use a row or column total.

SECTION 5

Types of Relative Frequency & Interpretation

Understanding when to use each type of relative frequency is just as important as computing it. The diagram below illustrates how data flows from the original table to each type of relative frequency, and the kind of question each one answers.

This flowchart guides you from a question's wording to the correct type of relative frequency. The key difference is always the denominator: the grand total for joint and marginal, or a row/column total for conditional.

Summary of relative frequency types and their interpretations

Type	Question Pattern	Denominator	Interpretation
Joint	"What percentage of all students are…?"	Grand total (N)	Proportion of entire data set in one specific cell
Marginal	"What percentage of all students prefer…?"	Grand total (N)	Overall distribution of one variable, ignoring the other
Conditional	"Of those who…, what percentage…?" or "Given that…"	Row or column total	Proportion within a specific sub-group

SECTION 6

Worked Example

A school cafeteria surveyed 250 students about their lunch choice (hot lunch or packed lunch) and their grade level (underclassmen: grades 9–10, or upperclassmen: grades 11–12). The results are shown below.

Cafeteria survey data

	Underclassmen	Upperclassmen	Total
Hot Lunch	90	60	150
Packed Lunch	40	60	100
Total	130	120	250

Finding All Three Types of Relative Frequency

Step 1 — Identify the Grand TotalThe grand total is found in the bottom-right corner of the table: N = 250. This is the denominator for joint and marginal relative frequencies.

N = 250

Step 2 — Compute a Joint Relative FrequencyQuestion: "What proportion of all students are underclassmen who eat hot lunch?" The cell count for (Hot Lunch, Underclassmen) is 90. Divide by the grand total: 90 ÷ 250 = 0.36. This means 36% of all surveyed students are underclassmen who eat hot lunch.

Joint RF = 90 ÷ 250 = 0.36 (36%)

Step 3 — Compute a Marginal Relative FrequencyQuestion: "What proportion of all students eat packed lunch?" The row total for Packed Lunch is 100. Divide by the grand total: 100 ÷ 250 = 0.40. So 40% of all students surveyed bring a packed lunch.

Marginal RF = 100 ÷ 250 = 0.40 (40%)

Step 4 — Compute a Conditional Relative FrequencyQuestion: "Of upperclassmen, what proportion eat packed lunch?" Now we restrict our view to only upperclassmen. The column total for upperclassmen is 120. The cell count for (Packed Lunch, Upperclassmen) is 60. Divide: 60 ÷ 120 = 0.50. Among upperclassmen specifically, 50% bring a packed lunch.

Conditional RF = 60 ÷ 120 = 0.50 (50%)

Step 5 — Interpret in ContextCompare the conditional relative frequencies: among underclassmen, the proportion who eat packed lunch is 40 ÷ 130 ≈ 0.308 (30.8%), while among upperclassmen it's 50%. This suggests that upperclassmen are more likely to bring a packed lunch than underclassmen. Such a comparison is exactly why conditional relative frequencies are so useful—they let us compare groups of different sizes on an equal footing.

Upperclassmen (50%) are more likely to bring packed lunch than underclassmen (30.8%)

SECTION 7

Strengths & Common Pitfalls

Key strengths and common pitfalls when working with relative frequencies

Strengths	Pitfalls
Allows fair comparisons between groups of unequal sizes	Using the wrong denominator (e.g., grand total when the question asks for a conditional)
Converts raw counts into intuitive percentages everyone can understand	Confusing correlation with causation — a pattern in relative frequencies does not prove one variable causes the other
Reveals patterns and potential associations between two categorical variables	Ignoring small sample sizes — a 100% rate based on 2 observations is very misleading
All relative frequencies in a complete table sum to 1.00, providing a built-in accuracy check	Rounding errors — when you round individual cells, they may not sum to exactly 100%

⚡ WATCH OUT FOR SIMPSON'S PARADOX

When you combine data from multiple groups, the overall trend can sometimes reverse. For example, a treatment might appear more effective overall but less effective within every subgroup. This is called Simpson's Paradox, and it reminds us to always examine conditional relative frequencies alongside joint ones. Think of it like averaging bowling scores: you might beat your friend overall because you bowled more games on easy lanes, even though your friend beat you on every lane.

SECTION 8

Connection to Probability & Advanced Statistics

Relative frequencies are actually the bridge between data tables and probability. When you compute a relative frequency, you're estimating a probability based on observed data. The table below shows how the concepts you've learned map directly to probability notation you'll encounter in more advanced courses.

How relative frequencies connect to formal probability

This Lesson	Probability Notation	Reads As…
Joint relative frequency of A and B	P(A ∩ B) or P(A and B)	The probability that both A and B occur
Marginal relative frequency of A	P(A)	The probability that A occurs (regardless of B)
Conditional relative frequency of A given B	P(A \| B)	The probability of A, given that B has occurred
All joint RFs sum to 1	ΣP = 1	Total probability is always 1

In future courses, you'll use these ideas to determine whether two variables are independent. Two variables are independent if P(A | B) = P(A)—that is, knowing B doesn't change the probability of A. You'll also encounter the chi-square test, which formally tests whether the differences between observed and expected relative frequencies are statistically significant. Everything starts with the skills you're building right now.

SECTION 9

Practice Problems

Use the following two-way table for all five problems. A fitness center surveyed 300 members about their preferred workout type and the time of day they typically exercise.

Fitness center survey data (300 members)

	Morning	Evening	Total
Cardio	90	60	150
Weights	45	105	150
Total	135	165	300

PROBLEM 1 — CONCEPTUAL

Explain in your own words why we use relative frequencies instead of just comparing raw counts. Give a real-life situation where raw counts alone would be misleading.

PROBLEM 2 — BASIC CALCULATION

What is the joint relative frequency of members who do cardio in the morning? Express your answer as a decimal and as a percentage.

PROBLEM 3 — INTERMEDIATE

Of the members who exercise in the evening, what proportion prefer weights? Identify which type of relative frequency this is and show your work.

PROBLEM 4 — APPLIED

The fitness center manager wants to decide whether to add more morning weightlifting classes. She reasons: "Since cardio and weights each have 150 members total, they're equally popular, so no changes are needed." Use conditional relative frequencies to explain why her reasoning is flawed, and what the data actually suggests about morning demand.

PROBLEM 5 — CRITICAL THINKING

If workout preference and time of day were independent (no association), what would the joint relative frequency of (Cardio, Morning) be? Compare this expected value to the observed joint relative frequency and discuss what the difference suggests about the relationship between the two variables.

SUMMARY

Lesson Summary

A two-way table organizes data by two categorical variables, and converting its raw counts into relative frequencies unlocks powerful comparisons. A joint relative frequency divides a cell count by the grand total and tells you the proportion of all observations in that cell. A marginal relative frequency divides a row or column total by the grand total and describes the overall distribution of one variable. A conditional relative frequency divides a cell count by its row or column total and answers "given that" questions about a specific subgroup.

The critical skill is choosing the correct denominator based on what the question asks. Use the grand total for joint and marginal frequencies; use a row or column total for conditional frequencies. Relative frequencies connect directly to probability and provide the foundation for testing whether two variables are independent. Always interpret your results in context—a number like 0.35 means nothing until you say "35% of all surveyed students prefer…"