Statistics › Section E · Continuous random variables › Topic E.1
Probability density functions
Discrete variables list values and attach probabilities; measured quantities take any value in an interval, and no list can hold them. For continuous random variables, probability becomes geometry: the chance of an interval is the area under a curve.
Probability becomes geometry
Discrete variables list values and attach probabilities to each. Measured quantities - times, weights, distances - take any value in an interval, and no list can hold them. For continuous random variables the whole framework changes shape: probability moves from table entries to geometry, and the chances of intervals are areas under a curve. This topic builds that curve, the probability density function, and the reading habits that go with it.
Why a single value has probability 0
A continuous variable can land anywhere on an interval, and its probability is spread over infinitely many values, so no single value can hold a positive share. The formal statement: \(P(X = c) = 0\) for any exact value \(c\). One immediate consequence: for continuous variables, \(P(X \le 3)\) and \(P(X < 3)\) are the same number, since the boundary point contributes nothing. Note what this does not mean: \(X = c\) is not impossible - it has probability 0 the way a single point has no width to carry area.
The density function f(x)
A continuous distribution is drawn as a curve \(f(x)\), the probability density function. Its rules: the curve never dips below the x-axis; the total area between curve and axis is exactly 1; and probability is assigned only to intervals, as the area above them. The height \(f(x)\) is a density - how concentrated probability is near \(x\) - not a probability itself. A taller stretch of curve means values there are likelier, the way a densely populated stretch of a map holds more people per mile.
Probability as area
Every probability question about a continuous variable is a question about area:
Shade the region above the interval, bounded by the curve, and find its area. Left-areas answer cumulative questions: \(P(X \le x)\) is everything up to \(x\), and as a function of \(x\) it is called the cumulative distribution function. Right-areas are complements: \(P(X > x) = 1 - P(X \le x)\), since the two regions partition the total area of 1.
Simple areas: rectangles and triangles
Without calculus, two shapes carry most examples. A constant density over \([a, b]\) - the uniform case - draws a rectangle of height \(\frac{1}{b-a}\), so that base times height gives area 1. A density rising linearly from zero draws a triangle, area \(\frac{1}{2} \times \text{base} \times \text{height}\). More irregular curves get their areas from tables or technology, as the normal-distribution chapter will show; the geometry here is the concept.
Worked example: the 10-minute bus
A bus arrives at some point uniformly within a 10-minute window, so \(f(x) = \frac{1}{10}\) on \([0, 10]\). a) \(P(X < 2)\): rectangle, base 2, height \(\frac{1}{10}\), area \(\frac{2}{10} = 0.2\). b) \(P(3 < X < 7)\): base 4, area 0.4. c) \(P(X = 5)\): a vertical line has no width, so the area - and the probability - is 0. d) Check: \(P(X < 2) + P(2 \le X \le 10)\) must total 1; the second rectangle has base 8 and area 0.8, and \(0.2 + 0.8 = 1\).
Worked example: a rising density
A service desk's waiting time \(X\), in hours, has density \(f(x) = \frac{x}{2}\) on \([0, 2]\): longer waits grow likelier up to the 2-hour cutoff. 1. Validity: the region is a triangle with base 2 and height \(f(2) = 1\), area \(\frac{1}{2} \times 2 \times 1 = 1\) - a legal density. 2. \(P(X \le 1)\): the triangle from 0 to 1 has height \(f(1) = \frac{1}{2}\), area \(\frac{1}{2} \times 1 \times \frac{1}{2} = 0.25\). 3. \(P(X > 1) = 1 - 0.25 = 0.75\). 4. Check: three-quarters of the area lies right of 1, matching the right-leaning triangular shape.
Reading the curve critically
Density curves invite a specific reading discipline. Locate the interval, shade its area, then compute. Ask whether the total area is 1 before trusting any curve as a distribution. Distinguish height from probability: \(f(5)\) can exceed 1 - a tall, narrow curve is legal so long as the total area stays 1 - whereas discrete probabilities are capped at 1. And remember the endpoints: including or excluding them changes nothing, because points carry no area.
Common mistakes to avoid
First, reading \(f(c)\) as \(P(X = c)\): height is not probability. Second, forgetting that total area must equal 1 - a curve enclosing 0.8 or 1.3 is not a distribution. Third, dropping the \(\frac{1}{2}\) from a triangle's area: base times height alone would make \(P(X \le 1)\) under \(f(x) = \frac{x}{2}\) equal 0.5, not 0.25. Fourth, using the wrong height: that triangle stops at \(x = 1\), so its height is \(f(1) = \frac{1}{2}\), not the curve's maximum \(f(2) = 1\). Fifth, expecting density values to stay below 1 as probabilities must. Sixth, applying discrete tools - point probabilities, table sums - to curves, where areas do the work.
Key takeaways
Continuous variables are measured, not counted, and their probability is geometry: the density \(f(x)\) never goes negative, encloses total area 1, and assigns probability to intervals as area. \(P(c < X < d)\) is the shaded area between \(c\) and \(d\); \(P(X = c) = 0\), so < and ≤ agree. Uniform densities draw rectangles of height \(\frac{1}{b-a}\); linearly rising densities draw triangles; both areas come from base times height. \(P(X \le x)\) is the left-area, and \(P(X > x) = 1 - P(X \le x)\). Height is density, never probability.
Flashcards
Flashcard
The graph of a continuous distribution
Test your understanding
Question 1 of 5
For a continuous random variable X, what is P(X = 5)?
Question 2 of 5
The total area under any probability density curve is:
Question 3 of 5
A bus arrives uniformly within a 10-minute window, f(x) = 1/10 on [0, 10]. What is P(X < 2)?
Question 4 of 5
A waiting time has density f(x) = x/2 on [0, 2]. What is P(X ≤ 1)?
Question 5 of 5
For a continuous random variable, how do P(X ≤ 3) and P(X < 3) compare?
