Measures of Center

⚡ Key Takeaways

Measures of center are single numbers that summarize the "typical" or "central" value of a dataset. The three classical measures are the mean (arithmetic average), median (middle value), and mode (most frequent value). They apply anywhere data must be summarized — surveys, economics, science, sports. Used since antiquity but formalized in statistics from the 17th–19th centuries.

💡 The Core Insight

A measure of center answers the question: "If I had to replace every value in this dataset with one single number, what should it be?" Each measure answers this differently because each minimizes a different kind of "total error" — the mean minimizes squared distance, the median minimizes absolute distance, and the mode minimizes the count of mismatches. Center is not one idea but a family of answers to one question.

DefinitionThe Arithmetic Mean

For a dataset of $n$ values $x_1, x_2, \dots, x_n$ , the arithmetic mean $\bar{x}$ is their sum divided by their count:

1

\quad \bar{x} = \frac{1}{n}\sum_{i=1}^{n} x_i = \frac{x_1 + x_2 + \cdots + x_n}{n}

Derivation hint: This is the value that, when multiplied by $n$ , recovers the total $\sum x_i$ — the "fair share" if the total were divided equally.

Symbols: $\bar{x}$ = sample mean; $n$ = number of data values; $x_i$ = the $i$ -th data value; $\sum$ = summation over all $i$ from $1$ to $n$ . For a whole population the mean is written $\mu$ .

DefinitionThe Median

Order the $n$ values from smallest to largest as $x_{(1)} \le x_{(2)} \le \cdots \le x_{(n)}$ . The median $M$ is the middle value:

2

\quad M = \begin{cases} x_{\left(\frac{n+1}{2}\right)} & n \text{ odd} \\[2mm] \dfrac{x_{\left(\frac{n}{2}\right)} + x_{\left(\frac{n}{2}+1\right)}}{2} & n \text{ even} \end{cases}

Derivation hint: For odd $n$ there is one exact middle position; for even $n$ the "middle" falls between two values, so we average the two straddling positions.

Symbols: $x_{(k)}$ = the $k$ -th value in sorted order (order statistic); $n$ = count; $M$ = median.

DefinitionThe Mode

The mode is the value (or values) that occurs most frequently. If $f(x)$ denotes the frequency of value $x$ , then

3

\quad \text{mode} = \arg\max_{x} f(x)

Derivation hint: " $\arg\max$ " returns the input $x$ at which the frequency function $f$ attains its maximum — the tallest bar in a frequency table. A dataset may have no mode (all distinct), one mode (unimodal), or several modes (bimodal, multimodal).

DefinitionThe Weighted Mean

When values carry different importances $w_i$ , the weighted mean is:

4

\quad \bar{x}_w = \frac{\sum_{i=1}^{n} w_i x_i}{\sum_{i=1}^{n} w_i}

Derivation hint: Treat each $x_i$ as if it appeared $w_i$ times; the numerator is the weighted total and the denominator is the total weight.

Symbols: $w_i$ = weight assigned to value $x_i$ . When all $w_i = 1$ , this reduces to equation [1].

TheoremThe Mean as a Balance Point

The arithmetic mean is the unique value $c$ for which the signed deviations sum to zero:

5

\quad \sum_{i=1}^{n}(x_i - \bar{x}) = 0

Derivation hint: Expand the sum into $\sum x_i - n\bar{x}$ , then substitute $\sum x_i = n\bar{x}$ from equation [1].

TheoremLeast-Squares Property of the Mean

The mean minimizes the total squared deviation. For any constant $c$ :

6

\quad \sum_{i=1}^{n}(x_i - c)^2 \text{ is minimized when } c = \bar{x}

Derivation hint: Differentiate $S(c)=\sum(x_i-c)^2$ with respect to $c$ , set $S'(c)=0$ , and solve — or complete the square.

ProofThe Mean Minimizes Squared Deviation

We prove equation [6] by calculus.

Step 1. Define the total squared deviation as a function of $c$ :

S(c) = \sum_{i=1}^{n}(x_i - c)^2.

Step 2. Differentiate with respect to $c$ using the chain rule:

S'(c) = \sum_{i=1}^{n} 2(x_i - c)(-1) = -2\sum_{i=1}^{n}(x_i - c).

Step 3. Set $S'(c) = 0$ to find the critical point:

\sum_{i=1}^{n}(x_i - c) = 0 \;\Longrightarrow\; \sum x_i - nc = 0 \;\Longrightarrow\; c = \frac{1}{n}\sum x_i = \bar{x}.

Step 4. Confirm it is a minimum: the second derivative is $S''(c) = 2n > 0$ , so the curve is upward-opening and the critical point is a minimum. Therefore the mean is exactly the value that minimizes total squared error. $\blacksquare$

∎

ProofThe Median Minimizes Absolute Deviation

We show the median minimizes $\sum |x_i - c|$ .

Step 1. Consider $A(c) = \sum_{i=1}^{n}|x_i - c|$ . For values $x_i > c$ the term contributes $+1$ to the slope as $c$ increases; for $x_i < c$ it contributes $-1$ .

Step 2. The slope of $A(c)$ is therefore (number of points below $c$ ) − (number above $c$ ).

Step 3. $A(c)$ stops decreasing and starts increasing exactly when the number of points on each side balances — i.e., at the middle value.

Step 4. That balancing point is the median (for even $n$ , any value between the two central points works, and convention takes their average). $\blacksquare$

∎

mathematician · Adolphe Quetelet | 1796–1874

The Belgian astronomer and statistician Adolphe Quetelet pioneered applying the arithmetic mean to human and social data, inventing the concept of l'homme moyen ("the average man"). He argued that physical and social traits cluster around averages following bell-shaped curves, founding social statistics. Though his "average man" was later criticized as a fiction, Quetelet transformed the mean from a tool of astronomers (averaging error-prone observations) into the cornerstone of demography, sociology, and modern data science. The Body Mass Index, originally the "Quetelet Index," still bears his legacy.

🧠 Intuition

Picture a long, weightless ruler with identical weights placed at each data value. The mean is the exact point where you could balance the ruler on a fingertip — move one weight far to the right and the balance point shifts toward it. This is why a single billionaire pulls the mean income upward dramatically. The median is different: imagine lining up everyone by height and pointing at the person standing dead center. Add a 9-foot giant to the line and the center person barely shifts — the median ignores how far outliers are, only how many are on each side. It is robust. The mode is simply the most crowded spot — the shoe size a store sells most, the most common exam score. It needs no arithmetic at all and is the only measure usable for categories like "favorite color."

Figure

Three measures of center on a right-skewed distribution

CF

Carl Friedrich Gauss

1777–1855

Gauss, studying errors in astronomical measurements, showed that the arithmetic mean is the most probable value when errors follow what we now call the Gaussian (normal) distribution. His method of least squares (1809) gave the mean a deep theoretical home: it is the estimate that minimizes squared error, exactly the property in equation [6]. This linked center-finding to probability and made the mean the default summary in the physical sciences.

ExampleQuiz Scores — All Three Measures

Given: Seven students score: $6, 8, 8, 9, 10, 10, 10$ (out of 10).

Mean (eq. [1]):

\bar{x} = \frac{6+8+8+9+10+10+10}{7} = \frac{61}{7} \approx 8.71.

Median (eq. [2]): $n = 7$ is odd; the middle position is $\frac{7+1}{2} = 4$ th value. Sorted data: $6,8,8,\mathbf{9},10,10,10$ .

M = 9.

Mode (eq. [3]): The value $10$ appears three times — more than any other.

\text{mode} = 10.

> Answer: mean $\approx \boxed{8.71 \text{ points}}$ , median $= \boxed{9 \text{ points}}$ , mode $= \boxed{10 \text{ points}}$ . The left-skew (low score of 6) pulls the mean below the median.

ExampleWeighted Course Grade

Given: A course grade weights Homework 20%, Midterm 30%, Final 50%. A student earns Homework $= 95$ , Midterm $= 80$ , Final $= 88$ .

Step 1 — apply the weighted mean (eq. [4]) with $w_1=0.20,\,w_2=0.30,\,w_3=0.50$ :

\bar{x}_w = \frac{(0.20)(95) + (0.30)(80) + (0.50)(88)}{0.20+0.30+0.50}.

Step 2 — numerator:

(0.20)(95)=19,\quad (0.30)(80)=24,\quad (0.50)(88)=44,\quad \text{sum}=87.

Step 3 — denominator $= 1.00$ , so:

\bar{x}_w = \frac{87}{1.00} = 87.

> Answer: Final grade $= \boxed{87 \text{ percent}}$ . Note the unweighted mean would be $\frac{95+80+88}{3}\approx 87.7$ — different, because the heavily-weighted Final dragged the result down.

⚠️ Common mistake

"The average is always the typical value." False. For skewed data — incomes, house prices, response times — the mean is pulled toward the tail and may exceed what most people actually have. If a town of 100 people includes one billionaire, the mean wealth might be $10 million while the median is $40,000. Always ask which measure and why before trusting a headline "average."

⚠️ Common mistake

"You can take the mean of anything numeric." Not meaningfully. The mean of jersey numbers, zip codes, or ranked categories (1=poor, 5=excellent) is often nonsensical because the numbers are labels or ordered ranks, not true quantities with equal spacing. Such ordinal or nominal data calls for the median or mode.

Counterexample

Mean fails badly with outliers. Dataset of reaction times (seconds): $0.3, 0.4, 0.3, 0.5, 0.4, 12.0$ — where $12.0$ is a recording error.

\bar{x} = \frac{0.3+0.4+0.3+0.5+0.4+12.0}{6} = \frac{13.9}{6} \approx 2.32 \text{ s}.

No actual reaction took anywhere near $2.32$ s! The median, $\frac{0.4+0.4}{2}=0.4$ s, faithfully represents typical performance. This single example shows why robustness matters.

connection

Measures of center never travel alone. They pair with Measures of Spread (range, variance, standard deviation) — a center without spread is half a story, since two datasets can share a mean yet differ enormously in consistency. The least-squares property (eq. [6]) directly defines the Variance and Standard Deviation, which measures squared deviation from the mean. The relationship between mean, median, and mode diagnoses Skewness: mean > median signals right-skew. In probability, these sample statistics estimate the population Expected Value $E[X]$ , and the median connects to Quartiles and Percentiles as the 50th percentile.

🔬 Where this appears

⏳ History

- c. 500 BCE — Pythagoreans study the arithmetic, geometric, and harmonic means as ratios in music and geometry.

- c. 4th century BCE — Greek astronomers informally average discordant observations to reduce error.

- 1500s–1600s — Navigators and astronomers (e.g., Tycho Brahe) systematically average repeated measurements at sea and sky.

- 1635 — Henry Gellibrand uses the mean of observations to estimate magnetic declination, an early documented scientific average.

- 1755 — Thomas Simpson formally argues that the mean of measurements is superior to a single observation, giving probabilistic justification.

- 1809 — Carl Friedrich Gauss publishes the method of least squares, proving the mean's optimality under normal error.

- 1835 — Adolphe Quetelet introduces l'homme moyen (the average man), launching social statistics.

- 1880s — Francis Galton popularizes the median and studies the distinction between center and spread.

- 1895 — Karl Pearson coins terms like "mode" and "standard deviation," formalizing the modern vocabulary of descriptive statistics.

open-problem

Which center is "right" is genuinely undecidable without a goal. Each measure minimizes a different loss (squared, absolute, or zero-one error), so the "best" measure depends entirely on what cost you wish to minimize — a value judgment, not a mathematical fact. Subtler issues persist: the mode is unstable for continuous data (it depends on how you bin values); the median is robust but throws away magnitude information; the mean is efficient but fragile. For multimodal or heavily skewed real-world data, no single number suffices, and statisticians increasingly favor reporting full distributions, multiple quantiles, or robust hybrid estimators (like trimmed means). The deep open question in applied statistics is not how to compute center but whether a single summary should ever be trusted — a debate sharpened by modern "big data," where averaging can erase exactly the rare patterns that matter.

Formal Statement

Why It Holds / Derivation

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also