Foundations — Gaussian - Normal distribution properties

Before you can read the parent note, you need to own every squiggle it writes. Below, each symbol is built from nothing: plain words → a picture → why the topic can't live without it. Read top to bottom; each row leans on the one above.

1. A random variable — the thing that wobbles

Picture it: imagine measuring the height of the next person who walks through a door. You know it'll be "around 170 cm" but not exactly. $X$ is the name of that not-yet-known number.

• A specific observed value gets a lowercase letter: $x$. So "$X=x$" means "the wobbly thing landed on the concrete value $x$."

Why the topic needs it: the whole Gaussian note is a description of how one particular kind of wobbly quantity wobbles.

2. Probability density $f(x)$ — the height of the curve

Here is the single most misunderstood symbol in the whole topic, so we build it slowly with a picture.

Picture it (look at the figure): the smooth hump is $f(x)$. Pick two points $a$ and $b$ on the horizontal axis. The shaded area between them, under the curve, is the probability that $X$ lands between $a$ and $b$.

Why the topic needs it: the parent's headline formula is $f(x)$. Every property (symmetry, the 68-95-99.7 rule, MLE) is a statement about this curve.

3. The integral $\int$ — "add up the area"

• $a$ (bottom of the sign) and $b$ (top) are the left and right edges of the region.
• $dx$ means "an infinitely thin slice of width $dx$."
• Each strip has area ≈ height $f(x)$ times width $dx$; the $\int$ sums them all.

Why this tool and not another? We need a total amount of area under a smooth curve. Ordinary addition can't add infinitely many strips; the integral is the exact tool invented for "continuous adding." That's the only reason it shows up whenever the parent writes $P(a\le X\le b)$.

4. Mean $\mu$ — where the peak sits

Picture it: slide the whole bell left or right; $\mu$ is the label on the axis under the tip.

Why the topic needs it: $\mu$ is one of the two numbers that completely pin down a Gaussian. Change $\mu$, the bell slides sideways without changing shape.

5. Variance ${\sigma }^2$ and standard deviation $\sigma$ — how wide

Picture it (look at the figure): the narrow blue bell has small $\sigma$ (values hug the center); the wide pink bell has large $\sigma$ (values spread out). Both share the same $\mu$, so both peak in the same place.

Why two names? We look at $\sigma$ (it's a width on the picture), but we compute with ${\sigma }^2$ (it adds when you add independent variables). The topic needs both hats.

6. The notation $X\sim \mathcal{N}(\mu ,{\sigma }^2)$

Why the topic needs it: it's the compact way of saying "this wobbly quantity's histogram is exactly the bell from Sections 2–5, tuned by these two numbers."

7. $\exp$ and $e$ — the smooth fall-off

Picture it: $\exp (0)=1$ (the peak), and as its input dives negative, the output slides smoothly toward zero — never touching it. That is exactly the graceful "tails that never quite hit the axis" shape of the bell.

Why this tool and not another? We need a curve that (a) is always positive (probabilities can't be negative), (b) peaks in the middle and falls off symmetrically, and (c) has the mathematically magic property that its area is computable in closed form. $e^{-(\text{squared distance})}$ is the unique friendly function with all three. That's why the parent's formula wraps a squared term inside $\exp$.

8. The squared distance $(x-\mu {)}^2$ — why the bell is symmetric

Picture it: a point 5 units left of $\mu$ and a point 5 units right of $\mu$ both give $(x-\mu {)}^2=25$. Since the curve's height depends only on this squared distance, the two points sit at the same height. That is precisely the parent's Property 1: symmetry.

9. The normalizing constant $\frac{1}{\sigma \sqrt{2\pi }}$

Why the topic needs it: without this constant, $f$ would be a valid shape but not a valid probability (its area wouldn't be 1). $\pi \approx 3.14159$ appears because the area of a bell famously involves $\pi$ — a deep fact you can accept for now.

10. $\Phi$, $z$, and $\text{erf}$ — the running-total of area

Picture it (look at the figure): the left shaded region is $\Phi (z)$. The tail to the right of $z$ is the leftover, $1-\Phi (z)$ — exactly the "$P(X>130)$" the parent computes in Example 1.

Why these tools? There is no elementary formula for the area under a Gaussian. So mathematicians gave that area a name ($\Phi$ / $\text{erf}$) and tabulated it once and for all. Every probability question in the parent ("what fraction beyond 130?", "68% within one $\sigma$?") is really a lookup of $\Phi$ or $\text{erf}$.

11. $\prod$, $\ln$, and $\frac{\partial }{\partial \mu }$ — the estimation toolkit

These appear in the parent's Maximum Likelihood section.

Why this tool and not another? "Best fit" means "highest likelihood." The top of any smooth hill is exactly where the slope is flat (zero). Derivatives are the machine that finds flat spots, so they're the natural tool for estimation.

Prerequisite map

Once every box above feels obvious, the parent note reads like plain English. From here the topic branches into Central Limit Theorem (why sums become Gaussian), Maximum Likelihood Estimation (fitting $\mu ,{\sigma }^2$ to data), and multi-dimensional cousins in the Multivariate Gaussian. Return to the parent: Gaussian - Normal distribution properties.

Equipment checklist

Test yourself — cover the right side and answer before revealing.