Skip to content

Lesson 19 ยท Random Variables

CDFs and Continuous Random Variables

Learn the cumulative distribution function, probability density functions, how probability becomes area under a curve, and compute probabilities for uniform and normal variables with SciPy.

A discrete random variable puts probability on separate points. A continuous one, such as a waiting time or a height, spreads probability smoothly along a line. Asking for P(X = 2.5) exactly is meaningless: among infinitely many values, any one has probability zero. Instead, we ask about ranges. This lesson introduces the two functions that describe that: the density and the cumulative distribution.

The cumulative distribution function

The cumulative distribution function (CDF) works for both discrete and continuous variables:

F(x) = P(X โ‰ค x)

It starts at 0 on the far left, never decreases, and ends at 1 on the far right. It answers "what share of outcomes lies at or below x?". Percentiles are its inverse: the 90th percentile is the x where F(x) = 0.9.

For a fair die, the CDF is a staircase: F(1) = 1/6, F(2) = 2/6, ..., F(6) = 1. For any value between integers it stays flat: F(3.7) = 3/6.

Probabilities of ranges come from differences: P(a < X โ‰ค b) = F(b) - F(a).

Python The CDF of a die

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The probability density function

For a continuous variable the probability density function (PDF) f(x) describes how probability is spread out. It is not a probability itself. Instead:

P(a โ‰ค X โ‰ค b) = area under f between a and b

A valid density satisfies f(x) โ‰ฅ 0 everywhere and a total area of 1. Heights of f can exceed 1, since only areas are probabilities. If the density near 2.5 is high, values near 2.5 are relatively likely; to find an actual probability multiply by an interval width.

The link to the CDF: F(x) is the area under f up to x, and f is the slope of F. That is F(x) = integral of f, and f(x) = F'(x). In terms of ideas: the CDF accumulates, the PDF is the rate of accumulation.

flowchart LR
    D["PDF f(x)<br/>density, height of the curve"] -->|"area up to x"| C["CDF F(x)<br/>P(X <= x)"]
    C -->|"slope"| D
    C --> P["P(a < X <= b) = F(b) - F(a)"]

Density and cumulative function carry the same information in two forms.

A first continuous example: the uniform distribution

If a bus arrives at a random moment within a 10-minute window, the waiting time X is equally likely anywhere in 0 to 10. The density is flat: f(x) = 1/10 on [0, 10]. Then

  • P(X โ‰ค 3) = 3 * 1/10 = 0.3
  • P(4 โ‰ค X โ‰ค 7) = 3/10 = 0.3
  • P(X = 5) = 0

The CDF is the straight line F(x) = x/10 between 0 and 10.

Python Uniform waiting time with SciPy

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

ppf is the percent point function (the inverse CDF); it returns percentiles.

The normal distribution, briefly

The most important continuous model is the bell-shaped normal distribution with mean mu and standard deviation sigma. Its density is a smooth symmetric curve. You do not integrate it by hand: tables or software supply the CDF. The empirical rule you already know is a statement about its CDF. The chapter on common distributions covers it fully.

Python Probabilities for a normal variable

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The probability within one standard deviation is about 0.683, matching the 68% rule. sf (survival function) is 1 - cdf and is more accurate for tail probabilities.

Plotting density and CDF together

Python PDF and CDF of the normal

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Expected value and variance for continuous variables

The definitions replace sums with integrals: E[X] = integral of x * f(x) dx, Var(X) = E[(X - mu)^2]. Every rule from the previous lesson, such as linearity, remains true. For the uniform distribution on [a, b]: mean (a + b)/2, variance (b - a)^2 / 12. For [0, 10] that gives 5 and 8.33.

Discrete versus continuous at a glance

Discrete Continuous
Described by PMF p(x) = P(X = x) PDF f(x) (density)
Probability of a single value Can be positive Always 0
Event probability Sum of PMF values Area under the PDF
CDF Staircase Smooth curve

Common mistakes

  • Reading a density value as a probability. Heights above 1 are fine.
  • Believing P(X = x) > 0 for a continuous variable.
  • Forgetting that P(a โ‰ค X โ‰ค b) and P(a < X < b) are equal for continuous variables.
  • Mixing up cdf (area to the left) and sf (area to the right).

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. For a continuous random variable, P(X = 3) equals...

    A single point has zero width, so zero area under the density.
  2. 2. X is uniform on [0, 20]. What is P(X <= 5)?

    The CDF is x/20, so F(5) = 0.25.
  3. 3. Which property is true of every CDF?

    F(x) = P(X <= x) accumulates probability, so it is non-decreasing from 0 to 1.
  4. 4. A density function can have height 4 at some point. Is that possible?

    Density is not probability. A narrow tall peak can still enclose an area of 1.

Summary

  • The CDF F(x) = P(X โ‰ค x) works for all random variables and gives range probabilities by subtraction.
  • Continuous variables are described by a density; probability is the area under it.
  • Single points have probability 0 for continuous variables.
  • The uniform and normal distributions are the first continuous models; SciPy provides pdf, cdf, sf and ppf.
  • The same ideas of mean and variance carry over, with integrals instead of sums.

Next: Joint Distributions and Covariance