Random Variables and Common Distributions
A random variable turns the outcome of a random experiment into a number (or vector of numbers). Distributions are the rulebook for those numbers. This chapter is the bridge from “probability of events” to “models we can fit and simulate.”
Diagram: the pipeline
sample space Ω
│ ω outcome
v
X(ω) = number ← random variable
│
├── discrete: PMF p(x) = P(X=x)
└── continuous: PDF f, CDF F
│
v
E[X], Var(X), quantiles, simulations
1. Discrete vs continuous
| Discrete | Continuous | |
|---|---|---|
| Values | countable | intervals of reals |
| Law | PMF \(p(x)\) | PDF \(f(x)\) with \(\int f=1\) |
| Probabilities | sum of PMF | areas under PDF |
| \(P(X=a)\) | may be positive | usually \(0\) for continuous |
CDF (always defined): \(F(x)=P(X\le x)\).
Properties of \(F\):
- non-decreasing
- right-continuous (standard construction)
- \(\lim_{x\to-\infty}F(x)=0\), \(\lim_{x\to+\infty}F(x)=1\)
For continuous \(X\) with PDF \(f\): \(F'(x)=f(x)\) where continuous, and
\[ P(a<X\le b)=F(b)-F(a)=\int_a^b f(t)\,dt. \]
2. Expectation and variance
Discrete:
\[ \mathbb{E}[X]=\sum_x x\,p(x),\qquad \mathrm{Var}(X)=\mathbb{E}[X^2]-(\mathbb{E}[X])^2. \]
Continuous: replace sum by integral against \(f\).
Linearity always: \(\mathbb{E}[aX+bY]=a\mathbb{E}[X]+b\mathbb{E}[Y]\) (no independence needed).
Variance of independent sum: \(\mathrm{Var}(X+Y)=\mathrm{Var}(X)+\mathrm{Var}(Y)\) if independent.
3. Catalog of workhorse distributions
Bernoulli(\(p\))
\(X\in\{0,1\}\), \(P(X=1)=p\).
\(\mathbb{E}[X]=p\), \(\mathrm{Var}(X)=p(1-p)\).
Binomial(\(n,p\))
Number of successes in \(n\) independent Bernoulli trials:
\[ P(X=k)=\binom{n}{k}p^k(1-p)^{n-k},\quad k=0,\ldots,n. \]
\(\mathbb{E}[X]=np\), \(\mathrm{Var}(X)=np(1-p)\).
trial: · · · · · · · · · · n flips
success? Y N Y Y N ...
X = count of Y
Geometric(\(p\)) (trials until first success)
\(\mathbb{E}[X]=1/p\) (if \(X\) counts trials). Model for retries until success.
Poisson(\(\lambda\))
Counts of rare events in a fixed window:
\[ P(X=k)=e^{-\lambda}\frac{\lambda^k}{k!}. \]
\(\mathbb{E}[X]=\mathrm{Var}(X)=\lambda\).
Exponential(\(\lambda\))
Waiting time between Poisson events; PDF \(f(x)=\lambda e^{-\lambda x}\) for \(x\ge 0\).
Memoryless: \(P(X>s+t\mid X>s)=P(X>t)\).
Normal(\(\mu,\sigma^2\))
\[ f(x)=\frac{1}{\sigma\sqrt{2\pi}}\exp\Big(-\frac{(x-\mu)^2}{2\sigma^2}\Big). \]
CLT: averages of i.i.d. noise often look approximately normal.
****
** **
** **
** **
-----*----------------------*----- x
μ-σ μ μ+σ
Uniform(\(a,b\))
Equal density on \([a,b]\); base case for inverse-transform sampling.
4. Worked examples
Example A. \(X\sim\mathrm{Binomial}(10,0.4)\).
\(\mathbb{E}[X]=4\), \(\mathrm{Var}(X)=10\cdot0.4\cdot0.6=2.4\).
Example B. \(X\sim\mathrm{Poisson}(5)\).
\(P(X=0)=e^{-5}\approx 0.0067\).
Example C (standardize). \(X\sim N(\mu,\sigma^2)\), then \(Z=(X-\mu)/\sigma\sim N(0,1)\).
\(P(X\le \mu+1.96\sigma)\approx 0.975\).
Example D (CDF from PMF). Support \(\{0,1,2\}\) with \(p=(0.2,0.5,0.3)\):
| \(x\) | \(F(x)\) |
|---|---|
| \(0\) | \(0.2\) |
| \(1\) | \(0.7\) |
| \(2\) | \(1.0\) |
5. Computer science map
| Model | CS use |
|---|---|
| Bernoulli / Binomial | A/B click, bit errors |
| Poisson | requests/sec, packet counts |
| Exponential | inter-arrival, simple reliability |
| Geometric | retry loops |
| Normal | measurement noise, CLT for metrics |
| Uniform | hash bucket idealization, RNG base |
6. How to choose a distribution (checklist)
- Discrete counts or continuous measurements?
- Bounded range or open-ended?
- Rare events / memoryless waiting?
- Symmetric noise around a mean?
- Fit with QQ-plot / histogram vs model — don’t only “pick normal by default.”
Exercises
- Derive \(\mathrm{Var}(X)\) for Bernoulli(\(p\)) from \(\mathbb{E}[X^2]-\mathbb{E}[X]^2\).
- For \(X\sim\mathrm{Binomial}(20,0.5)\), find \(\mathbb{E}[X]\) and \(\mathrm{Var}(X)\). Is \(P(X=10)\) larger or smaller than \(P(X=0)\)? Why?
- Build the CDF of a discrete RV with PMF \(p(-1)=p(1)=1/4\), \(p(0)=1/2\). Sketch the step function (ASCII is fine).
- If failures arrive as Poisson(\(3\)) per hour, approximate \(P(\text{at least one failure in 20 minutes})\).
- Explain in one paragraph when a normal approximation to Binomial(\(n,p\)) is poor.
- Challenge: Show memorylessness of Exponential: \(P(X>s+t\mid X>s)=P(X>t)\).
Quick answers (check after you try)
- \(p-p^2=p(1-p)\).
- Mean \(10\), var \(5\); \(P(X=10)\gg P(X=0)\).
- Window \(1/3\) hour → \(\lambda=1\), \(P(\ge 1)=1-e^{-1}\).
- When \(np\) or \(n(1-p)\) is small, or \(n\) tiny — mass piles near boundary.