Statistics Fundamentals

Statistics is the grammar of science. Whether you're analysing experimental data, building machine-learning models, or simply reading a poll, a handful of core ideas are all you need to make sense of numbers. This article covers the essentials — with interactive simulations to build real intuition.

---

1. Central Tendency — Where Is the Data?

When you have a dataset, the first question is: what is a typical value? Three measures answer this differently.

Mean — the arithmetic average. Sum all values and divide by $n$:

$\bar{x} = \frac{1}{n}\sum_{i=1}^{n} x_i$

The mean is sensitive to outliers. One extreme value can pull it far from the bulk of the data.

Median — the middle value when the data is sorted. If $n$ is even, it's the average of the two middle values. The median is robust: a single outlier can't move it much. This is why economists report median household income rather than mean — a handful of billionaires would inflate the mean beyond any useful meaning.

Mode — the most frequently occurring value. Useful for categorical data (e.g., most popular shoe size) and for identifying peaks in a distribution.

Try it: Switch between Symmetric, Right-skewed, and Left-skewed distributions. Notice how the mean chases the long tail while the median stays closer to the centre of mass.

---

2. Standard Deviation — How Spread Out Is the Data?

The mean tells you where the data sits. The standard deviation ($\sigma$) tells you how spread out it is around that centre.

$\sigma = \sqrt{\frac{1}{n}\sum_{i=1}^{n}(x_i - \bar{x})^2}$

The formula first computes the average squared distance from the mean (the variance, $\sigma^2$), then takes the square root to return to the original units.

A small $\sigma$ means values cluster tightly around the mean. A large $\sigma$ means they're spread wide. On a normal distribution, increasing $\sigma$ flattens and widens the bell curve — but the total area underneath always stays exactly 1 (it's a probability distribution).

Try it: Set σ₁ = 1 and σ₂ = 4. The wider curve has the same centre and the same total area — it just distributes probability over a larger range.

---

3. The Empirical Rule — 68–95–99.7

For a normal distribution (the symmetric bell curve), the proportion of data within each σ band follows a remarkably clean pattern:

Range	Approx.	Exact
$\mu \pm 1\sigma$	~68%	68.27%
$\mu \pm 2\sigma$	~95%	95.45%
$\mu \pm 3\sigma$	~99.7%	99.73%

This is sometimes called the 68-95-99.7 rule or the three-sigma rule. It shows up everywhere:

• In manufacturing, "six sigma" quality means fewer than 3.4 defects per million (±6σ from the target).
• In particle physics, a "5σ discovery threshold" means the probability of seeing the result by chance is less than 1 in 3.5 million.
• In finance, a "2σ event" in asset returns is expected to happen roughly 1 in 20 trading days.

Try it: Toggle between 1σ, 2σ, and 3σ ranges. The annotation shows the exact percentage of the distribution contained in each band.

---

4. Z-Scores — Standardising Any Value

A Z-score converts any measurement into a universal scale: how many standard deviations from the mean is this value?

$z = \frac{x - \mu}{\sigma}$

This lets you compare values from completely different scales. A student scoring 72 on a test with $\mu = 65$, $\sigma = 5$ has $z = 1.4$ — moderately above average. A patient with a blood pressure reading can be standardised to the same scale for comparison.

The colour indicator in the simulation follows a common convention:

• $|z| < 1$: typical — most values land here (68% of the distribution)
• $|z| \geq 1$: moderate
• $|z| \geq 2$: unusual — only ~5% of values are this far out
• $|z| \geq 3$: extreme — less than 0.3% probability under a normal distribution

Try it: Move the value $x$ while keeping μ and σ fixed. Watch the z-score and classification update in real time.

---

5. Confidence Intervals — Estimating the Unknown

In practice, you rarely know the true population mean $\mu$. You draw a sample and compute its mean $\bar{x}$. The question is: how close is $\bar{x}$ to $\mu$?

A confidence interval gives a plausible range for the true parameter:

$\text{CI} = \bar{x} \pm z^* \cdot \frac{\sigma}{\sqrt{n}}$

where:

• $\sigma / \sqrt{n}$ is the standard error (SE) — the standard deviation of the sample mean
• $z^*$ is the critical value: 1.645 for 90%, 1.96 for 95%, 2.576 for 99% confidence

A common misconception: "95% confidence" does not mean there's a 95% chance the true mean falls in this particular interval. The true mean either is or isn't in the interval — it's fixed. What it means is: if you repeated the experiment many times and built a CI each time, 95% of those intervals would contain the true mean.

Two levers narrow the interval:

1. Increase $n$ (more data) — SE shrinks as $1/\sqrt{n}$
2. Decrease confidence level — a 90% CI is narrower than a 99% CI, but less reliable

Try it: Drag the sample size slider from 5 to 500. The interval collapses as data accumulates.

---

Interactive Simulations

---

Quick Reference

Concept	Formula	What it tells you
Mean	$\bar{x} = \sum x_i / n$	Arithmetic average
Median	Middle value (sorted)	Robust centre (outlier-resistant)
Std Deviation	$\sigma = \sqrt{\sum(x_i-\bar{x})^2/n}$	Spread in original units
Z-score	$z = (x - \mu)/\sigma$	Standard deviations from the mean
Standard Error	$SE = \sigma/\sqrt{n}$	Uncertainty of the sample mean
95% CI	$\bar{x} \pm 1.96 \cdot SE$	Plausible range for the true mean

Further Reading

• Khan Academy — Statistics and Probability
• 3Blue1Brown — But what is a normal distribution?
• Seeing Theory — Brown University