PDF vs. CDF: Two Ways to Describe the Same Distribution

Every distribution covered elsewhere on this site — normal, binomial, exponential, and the rest — can be described in two complementary ways: how likely each specific value (or narrow range of values) is, and how much total probability has accumulated up to a given point. These are the probability density function (PDF) and the cumulative distribution function (CDF), and confusing the two is a common source of errors when reading distribution formulas or software output.

The Core Distinction

The CDF, denoted F(x), gives the probability that a random variable is less than or equal to a specific value x — it always ranges from 0 to 1 and is always non-decreasing, since accumulated probability can never shrink as x increases. The PDF, denoted f(x), describes the relative likelihood of the variable taking values near x — for a continuous distribution, the PDF's value at any single exact point isn't itself a probability (a continuous variable has zero probability of landing on any exact value), but the area under the PDF curve between two points equals the probability of falling in that range, which is exactly what the CDF calculates directly.

Worked Example: The Standard Normal Distribution

Problem: For the standard normal distribution, find both the PDF and the CDF at z=1.5, and explain what each number means.

PDF: φ(1.5) ≈ 0.1295. This is the height of the bell curve at z=1.5 — useful for comparing relative likelihood between points (the curve is roughly 0.13 units "tall" there) but not itself a probability.

CDF: Φ(1.5) ≈ 0.9332. This is the actual probability that a standard normal variable is less than or equal to 1.5 — the same 93.32% figure used throughout normal distribution and z-score calculations elsewhere on this site whenever "the probability of being below this point" is the question being asked.

The key practical difference: whenever a genuine probability is needed — P(X≤x), P(X>x), or P(a≤X≤b) — the CDF is what actually answers it (directly, or through simple subtraction). The PDF's role is different: it's the mathematical object that gets integrated (summed, in the continuous calculus sense) to produce the CDF in the first place, and it's what's plotted to visualize a distribution's overall shape.

A Cleaner Contrast: The Discrete Case

Problem: For a binomial distribution with n=5 trials and p=0.5, find the PMF (the discrete equivalent of a PDF) at X=3, and the CDF at X=3.

PMF: P(X=3) = C(5,3)(0.5)³(0.5)² = 0.3125 — this genuinely is a probability (the probability of exactly 3 successes), unlike the continuous PDF's height value. This is the one place the discrete and continuous cases differ conceptually: a discrete distribution's "PDF" (properly called a probability mass function, or PMF) gives real probabilities directly at each individual value, since a discrete variable genuinely can land on one exact value with positive probability.

CDF: F(3) = P(X≤3) = P(0)+P(1)+P(2)+P(3) = 0.8125 — the cumulative probability of getting 3 or fewer successes, found by summing the PMF values up to and including X=3.

Notice the CDF value (0.8125) is always at least as large as the corresponding PMF value (0.3125) at the same point, since the CDF is accumulating everything up to and including that point, while the PMF is isolating just that one specific value.

Reading the Relationship Graphically

A PDF or PMF plotted on its own shows a distribution's characteristic shape — a bell curve, a skewed hump, discrete bars of varying height. The corresponding CDF, plotted separately, always looks like a non-decreasing curve starting at 0 (or near it) and rising to 1 — for a continuous distribution, a smooth S-shaped curve; for a discrete distribution, a step function that jumps up at each possible value by exactly that value's PMF probability. Where the PDF is tall (a likely region), the CDF rises steeply; where the PDF is near zero (an unlikely region), the CDF stays nearly flat.

Real-World Uses

Software and calculator functions: nearly every statistical software package and calculator on this site offers both a density/mass function and a cumulative function for each distribution, precisely because different questions call for different ones — recognizing which is which from a function's name or documentation avoids a common source of confusion when reading unfamiliar statistical software output.

Risk and reliability analysis: engineers and actuaries routinely need cumulative probabilities (the chance a component fails before a certain time, the chance a claim exceeds a certain amount) — exactly what the CDF is built to answer directly, while the PDF is used more for visualizing where risk concentrates.

Percentile and quantile calculations: any time a specific percentile is needed — the 95th percentile cutoff for a control chart, the median of a fitted distribution — the calculation works from the CDF, inverted to solve for the x-value at a target cumulative probability, rather than from the PDF directly.

Common Mistakes

Treating a continuous PDF's height as a probability. A PDF value of, say, 1.2 is entirely possible and doesn't violate anything — PDF values aren't capped at 1 the way probabilities are, since they're densities, not probabilities; only the area under the curve (computed via the CDF) is a genuine probability.

Forgetting that P(X=x) is exactly 0 for a continuous variable. This can feel counterintuitive, but it's a direct consequence of a continuous variable's infinitely many possible values — meaningful continuous probabilities are always about ranges (P(a≤X≤b)), never single exact points, which is why "P(X=1.5) for a continuous normal variable" is a different, always-zero question from "the PDF's height at 1.5."