Random Variables and Distributions

PDF, CDF, and the Survival Function - the building blocks of probability distributions.

Before we can understand fat tails, we need a solid foundation in probability. This section introduces the key mathematical objects: random variables and their distributions.

What is a Random Variable?

Definition

Random Variable

A random variable is a function that assigns numerical values to outcomes of a random experiment. We write where is the sample space (all possible outcomes) and is the set of real numbers.

There are two main types of random variables:

  • Discrete: Takes countable values (coin flips, dice rolls, number of customers)
  • Continuous: Takes values in an interval (heights, stock returns, time)

For Taleb's work on fat tails, we focus primarily on continuous random variables, though the concepts apply to both types.

The Probability Density Function (PDF)

Definition

Probability Density Function

For a continuous random variable , the probability density function describes the relative likelihood of taking on different values. It satisfies:

  1. for all
  2. (total probability is 1)
Example

Normal (Gaussian) Distribution PDF

The famous bell curve — also called the Gaussian distribution or simply the normal distribution — has PDF:

Read: f of x equals one over sigma root two pi, times e to the minus x minus mu squared over two sigma squared

The height of the bell curve at point x, centered at μ with spread σ

The standard normal is the special case where (centered at zero) and (unit spread). It's often written as .

Interactive: Probability as Area Under the Curve

Probability as area: P(a ≤ X ≤ b) = area under the curve between a and b
μmean
0.00
σstd dev
1.00

Select range [a, b]

alower bound
-1.00
bupper bound
1.00
Loading chart...

P(-1.00 ≤ X ≤ 1.00)

74.07%

= F(1.00) − F(-1.00) = 0.8703 0.1297

Quick presets (Normal distribution)

Probability as area: This visualization shows the fundamental relationship between the PDF and probability. The shaded area under the curve between a and b equals P(a ≤ X ≤ b). This is calculated using the CDF: P(a ≤ X ≤ b) = F(b) − F(a).

The Cumulative Distribution Function (CDF)

Definition

Cumulative Distribution Function

The CDF gives the probability that is less than or equal to :

Key properties of the CDF:

  • is always non-decreasing (never goes down)
  • The PDF is the derivative of the CDF:
Key Insight

From PDF to CDF and Back

The CDF is the cumulative (running total) area under the PDF curve. Conversely, the PDF is the slope of the CDF. These two functions contain the same information in different forms.

Explore: PDF and CDF

Use this interactive tool to see how the PDF and CDF relate. Adjust parameters and see how the shaded area under the PDF corresponds to the CDF value.

μmean
0.00
σstd dev
1.00

Probability Density Function (PDF)

Loading chart...

Cumulative Distribution Function (CDF)

Loading chart...
xvalue
0.00

P(X ≤ 0.00) = F(0.00) = 0.5000

The Survival Function

Definition

Survival Function

The survival function (also called the tail function or complementary CDF) is:

Read: S of x equals the probability that X exceeds x, which equals one minus F of x

What fraction of the probability mass lies above the value x?

Key Insight

Why the Survival Function Matters for Fat Tails

The survival function tells us how much probability remains in the tails. For fat-tailed distributions, decays slowly — extreme events remain relatively probable even far from the mean.

Taleb's central observation: the difference between distributions is most visible in their survival functions, especially at large values of .

Compare how different distributions behave in the tails:

  • Normal: — decays exponentially fast
  • Exponential: — decays exponentially
  • Pareto (power law): — decays much slower!

Preview: Gaussian vs Pareto

Before we dive deeper, here's a visual comparison. The Pareto distribution is the archetypal "fat-tailed" distribution — we'll study it in detail in Module 2.

Gaussian (Normal) — symmetric bell curve

Loading chart...

Pareto — heavy right tail

Loading chart...
αtail exponent
2.50

Lower α = heavier tail (more probability of extreme values). The Pareto only exists for x ≥ 1 (its minimum value).

Notice how the Gaussian drops off quickly and symmetrically, while the Pareto has a long, slowly-decaying right tail. This "heavy tail" is what makes extreme events more likely under Pareto-like distributions.

Explore: Survival Functions and Tail Behavior

This plot compares how quickly different distributions decay in the tails. Toggle log-log scale to see the characteristic straight-line behavior of power law distributions.

αtail exp.
2.00
Loading chart...

Key insight: On a log-log plot, the Pareto distribution appears as a straight line, while the normal and exponential curves drop off much faster. This straight line is the signature of a power law or fat tail — extreme events remain relatively probable even far from the mean.

Notice: With α ≤ 2, the Pareto distribution has infinite variance. With α ≤ 1, even the mean is infinite!

Key Takeaways

  • A random variable assigns numbers to random outcomes
  • The PDF describes relative likelihoods; area under the curve gives probability
  • The CDF gives cumulative probability
  • The survival function measures tail probability — this is where fat tails reveal themselves
  • The rate at which decays determines whether a distribution is thin-tailed or fat-tailed