Random Number Generator

What Is a Random Number Generator?

A random number generator produces numbers from a specified probability distribution. Whether you need uniform random numbers between 1 and 100, normally distributed values with mean 50 and SD 10, or exponentially distributed waiting times, this tool generates samples from various distributions.

Supported Distributions

Uniform Distribution

Every value in a range is equally likely.

Parameters: Minimum, Maximum Uses: Random selection, simulation initialization, lottery numbers

Example: Uniform(1, 100) might produce: 47, 83, 12, 91, 34...

Normal (Gaussian) Distribution

The classic bell curve.

Parameters: Mean (μ), Standard Deviation (σ) Uses: Heights, test scores, measurement errors, many natural phenomena

Example: Normal(100, 15) might produce: 112, 87, 103, 91, 108...

Exponential Distribution

Time between events in a Poisson process.

Parameters: Rate (λ) or Mean (1/λ) Uses: Waiting times, service times, radioactive decay, component lifetimes

Example: Exponential(rate=0.5) might produce: 1.2, 0.3, 2.8, 0.7, 1.5...

Binomial Distribution

Number of successes in n trials.

Parameters: n (trials), p (success probability) Uses: Quality control, election modeling, clinical trials

Example: Binomial(n=10, p=0.3) might produce: 4, 2, 3, 1, 5...

Poisson Distribution

Count of events in fixed time/space.

Parameters: λ (average rate) Uses: Call arrivals, defect counts, rare events

Example: Poisson(λ=4) might produce: 3, 6, 2, 5, 4...

Applications

Monte Carlo Simulation

Random numbers drive simulations:

• Financial risk modeling
• Engineering reliability analysis
• Physics particle simulations
• Game development

Statistical Sampling

Select random samples from populations:

• Survey sampling
• Quality control inspection
• Clinical trial randomization

Hypothesis Testing

Generate null distributions:

• Permutation tests
• Bootstrap confidence intervals
• Power analysis via simulation

Teaching and Learning

Demonstrate statistical concepts:

• Central Limit Theorem
• Law of Large Numbers
• Distribution shapes

Games and Lotteries

Fair random selection:

• Dice simulation
• Card shuffling
• Prize drawings

True Random vs. Pseudo-Random

Pseudo-Random Number Generators (PRNGs)

Computer algorithms that produce sequences that appear random:

• Deterministic (reproducible with same seed)
• Fast and efficient
• Sufficient for most statistical applications

Common algorithms: Mersenne Twister, PCG, Xorshift

True Random Number Generators

Based on physical phenomena:

• Radioactive decay
• Atmospheric noise
• Quantum effects

Used for cryptography and high-stakes applications.

For statistical simulation, PRNGs are typically sufficient.

Setting Seeds

A seed initializes the random sequence:

• Same seed → same sequence
• Different seeds → different sequences

Why use seeds:

• Reproducibility: Re-run simulations exactly
• Debugging: Isolate problems
• Documentation: Share exact data generation

Distribution Transformations

Uniform to Any Distribution

The inverse transform method:

1. Generate U ~ Uniform(0, 1)
2. Apply inverse CDF: X = F⁻¹(U)

This converts uniform random numbers to any distribution with known inverse CDF.

Box-Muller Transform

Converts uniform to normal:

1. Generate U₁, U₂ ~ Uniform(0, 1)
2. Z₁ = √(-2ln(U₁)) × cos(2πU₂)
3. Z₂ = √(-2ln(U₁)) × sin(2πU₂)

Produces two independent standard normal values.

Quality Checks for Random Numbers

Good random sequences should:

• Have correct distribution shape
• Show no serial correlation
• Pass randomness tests (e.g., Diehard tests)

For most applications, standard library generators are adequate.

Historical Context

John von Neumann (1940s) developed early computational random number methods.

His "middle-square" method was simple but flawed. Better algorithms followed.

The Mersenne Twister (1997) by Makoto Matsumoto and Takuji Nishimura became the standard for statistical applications—long period (2^19937 - 1) and good statistical properties.

Modern languages include high-quality PRNGs by default.

Sample Size Considerations

For simulation:

• Small samples (n < 100): High variability between runs
• Moderate samples (n = 1000): Reasonable stability
• Large samples (n > 10000): Estimates approach theoretical values

More random numbers = more precise simulation results.

Stratified Random Sampling

Instead of pure random sampling:

1. Divide population into strata
2. Randomly sample within each stratum

Reduces variance compared to simple random sampling.

Common Mistakes

Reusing seeds unintentionally: Every simulation run uses the same sequence—no variance assessed.

Small sample sizes: Simulations with few random numbers have high Monte Carlo error.

Wrong distribution: Using normal when exponential is appropriate distorts results.

Related Calculations

• Normal Probability Calculator — Normal distribution
• Binomial Probability Calculator — Binomial distribution
• Poisson Probability Calculator — Poisson distribution
• Mean Calculator — Analyze generated samples
• Standard Deviation Calculator — Verify variance

Summary

Random number generators produce samples from specified probability distributions—essential for simulation, sampling, and demonstrating statistical concepts. From uniform to normal to Poisson, each distribution models different real-world phenomena. Pseudo-random generators provide reproducible sequences suitable for most statistical applications, while setting seeds enables exact replication of simulations. Whether running Monte Carlo simulations, conducting bootstrap analyses, or demonstrating the Central Limit Theorem, random number generation is fundamental to modern statistics.