Mean, Median, Mode

Why do we need “central tendency”?

Data sets can be long lists of numbers: test scores, delivery times, heights, or daily temperatures. Looking at a whole list makes it hard to compare groups or describe what’s normal.

A measure of central tendency compresses a data set into a single representative value. But “representative” can mean different things:

• •Balance point of the data (mean)
• •Middle of the data (median)
• •Most common value (mode)

These are not interchangeable; they answer different questions.

The three definitions (with intuition)

Suppose we have a data set of n values:

x₁, x₂, …, xₙ

1) Mean (arithmetic mean)

• •Idea: Share the total equally among all data points.
• •Symbol: x̄ (“x-bar”) for the sample mean.
• •Interpretation: If the data were weights on a number line, the mean is the balance point.

2) Median

• •Idea: The “middle” observation after sorting.
• •Interpretation: Half the data are ≤ median, and half are ≥ median.

3) Mode

• •Idea: The most frequent value(s).
• •Interpretation: What value shows up the most often?

When they agree—and when they don’t

If a distribution is perfectly symmetric and has no extreme outliers, mean ≈ median ≈ mode.

But with skew or outliers:

• •The mean is pulled toward extreme values.
• •The median stays near the center of the ordered list.
• •The mode depends only on frequency.

Quick comparison table

Important note about data types

• •Mean and median generally require numeric data.
• •Mode works for numeric or categorical data (e.g., most common color).

Why the mean is useful

The mean answers: “If I had to replace the whole data set with a single number that preserves the total, what number would that be?”

If you sum all values and then distribute that sum equally across n items, each would get the mean.

Definition

For values x₁, x₂, …, xₙ, the sample mean is

x̄ = (1/n) ∑ᵢ₌₁ⁿ xᵢ

Where:

• •n is the number of data points
• •∑ (“Sigma”) means “add up a bunch of things”

What ∑ means (unpacking the notation)

The expression

∑ᵢ₌₁ⁿ xᵢ

means:

x₁ + x₂ + x₃ + … + xₙ

So the mean is:

x̄ = (x₁ + x₂ + … + xₙ) / n

A key property: the mean preserves the total

If x̄ is the mean, then

n · x̄ = ∑ᵢ₌₁ⁿ xᵢ

This is a simple but powerful fact: using the mean keeps the sum consistent.

Mean as a balance point (intuition)

Imagine each data value xᵢ as a weight placed on a number line. The mean is where the system balances.

Another way to say this uses deviations from the mean:

Let dᵢ = xᵢ − x̄.

Then the deviations sum to zero:

∑ᵢ₌₁ⁿ (xᵢ − x̄) = 0

Show the work:

∑ᵢ₌₁ⁿ (xᵢ − x̄)

= ∑ᵢ₌₁ⁿ xᵢ − ∑ᵢ₌₁ⁿ x̄

= ∑ᵢ₌₁ⁿ xᵢ − n·x̄

= ∑ᵢ₌₁ⁿ xᵢ − ∑ᵢ₌₁ⁿ xᵢ

= 0

This is one reason the mean appears everywhere in statistics and machine learning: it’s the “center” that makes positive and negative deviations cancel.

The mean and outliers

Because the mean uses every value, one extreme point can change it a lot.

Example intuition:

• •Data: 10, 10, 10, 10
• •Mean = 10
• •Add an outlier 1000:
• •New mean = (10+10+10+10+1000)/5 = 208

Even if 4 out of 5 values are still 10, the mean shifts dramatically.

When the mean is the right tool

Use the mean when:

• •Outliers are rare or not meaningful
• •You care about totals (e.g., average revenue)
• •The distribution is roughly symmetric

Be cautious when:

• •There are extreme values (income, house prices)
• •The distribution is strongly skewed

Why median and mode exist

Sometimes “typical” should not be affected by extremes.

• •If one billionaire moves into a neighborhood, the mean income can jump a lot.
• •But the median income usually barely changes.

Likewise, for categorical data (like shirt sizes), mean may not even make sense—but mode still does.

Median

Definition

1) Sort the data from smallest to largest.

2) If n is odd: the median is the single middle value.

3) If n is even: the median is the average of the two middle values.

Let the sorted values be:

x₍₁₎ ≤ x₍₂₎ ≤ … ≤ x₍ₙ₎

• •If n is odd, median = x₍(n+1)/2₎
• •If n is even, median = (x₍n/2₎ + x₍n/2+1₎)/2

Why sorting matters

Median depends on order, not the actual magnitudes beyond their relative position.

So if you change an extreme value (like the max) but it remains at the far end, the median often stays the same.

Robustness to outliers

Because median is determined by the middle position(s), outliers have limited influence.

Example idea:

• •Data: 2, 3, 4, 5, 1000
• •Median is 4 (the middle)
• •Even though 1000 is huge, it does not pull the median upward.

When the median is the right tool

Use the median when:

• •You want a “typical” value in skewed distributions (income, time-to-complete)
• •Outliers are meaningful but shouldn’t dominate the summary

Mode

Definition

The mode is the value (or values) that occur most frequently.

• •If one value occurs most frequently: unimodal
• •If two values tie: bimodal
• •If many: multimodal
• •If all values occur equally often: sometimes said to have no mode

Why mode is useful

Mode answers: “What is the most common outcome?”

This is especially useful for:

• •Categorical data (most common color, most common major)
• •Discrete numeric data (most common number of siblings)

Mode can be “messy”

Mode can be unstable if:

• •The data are continuous with many unique values (e.g., exact measured heights)
• •Small sample sizes create accidental ties

In practice, for continuous data we often bin values (histograms) and talk about the most common range rather than the exact value.

Putting mean, median, mode together

A common pattern in right-skewed data (long tail to the right) is:

mode ≤ median ≤ mean

Because the mean is pulled toward the tail.

In left-skewed data:

mean ≤ median ≤ mode

These are not laws, but good intuition checks.

The real skill: choosing what “center” should mean

In practice, computing mean/median/mode is easy. The hard part is choosing the right one for the question and the data.

Common scenarios

1) Salaries / wealth (skew + outliers)

• •A few extremely high values
• •Mean can be misleading as a “typical” salary
• •Median is often reported (median household income)

2) Delivery times / latency (skew + occasional spikes)

• •Most deliveries are fast; some are delayed heavily
• •Median gives a typical experience
• •Mean captures the impact of delays (important for cost)

Often teams report both:

• •Median latency (user experience)
• •Mean latency (overall performance/cost)

3) Test scores in a class (often roughly symmetric)

• •Mean works well if no extreme anomalies
• •Median is a backup if one student has an unusual score due to absence/cheating/etc.

4) Product sizes / categories

For categorical data like “small/medium/large,” mean is not meaningful.

• •Mode tells you the most common size ordered.

A decision table

Connection to later statistics

Mean, median, and mode are the first step toward:

• •Variance/standard deviation (spread around the mean)
• •Quantiles/percentiles (generalizing the median)
• •Distributions (shape explains why mean and median differ)

If you remember one guiding idea: mean uses magnitudes, median uses order, mode uses counts.