Can a density exceed one?

Yes; only its integrated probability over a set must lie between zero and one.

Are recorded measurements truly continuous?

Instruments have finite precision, but a continuous model can still be a useful approximation.

Can a distribution combine a point mass and a continuous part?

Yes. Such a mixed distribution needs both components, for example a positive probability of zero spending plus varying positive amounts.

AI Glossary / Statistics and probability / Intermediate

Continuous probability distribution

Definition

A continuous probability distribution describes uncertainty in a quantity whose probability is spread across a continuum, rather than concentrated on separate values with positive probability. In the density-based distributions commonly used in statistics, probabilities are areas under a probability density function. A particular exact value has probability zero, while an interval can have positive probability.

What it means

A delivery duration, component length, or measurement error can be modeled on a continuous scale. The distribution describes how likely different ranges are, including the center, spread, asymmetry, and tails. It is more informative than reporting only an average.
Continuous does not mean normally distributed. Uniform, exponential, gamma, and many other distributions are continuous models with different shapes and supports. The model must match the kind of quantity and the question. A normal model can assign probability to negative values, which may be unsuitable for some strictly positive measurements.

How it works

Specify the variable, its units, and the values it can plausibly take. Choose or estimate a distribution using substantive knowledge and data. For a distribution with density f, the density is nonnegative and its total area is one. The probability of an interval is the integral of f over that interval.
The cumulative distribution function F(x) gives P(X <= x). Interval probabilities can be calculated as F(b) - F(a), and quantiles reverse this relationship to identify thresholds associated with specified cumulative probabilities. Simulation draws values according to the distribution and can help evaluate decisions under uncertainty.
A density height is not itself a probability. Its units are inverse measurement units, and it can exceed one. Changing from minutes to seconds changes the density's numerical height while leaving probabilities for equivalent physical intervals unchanged.
Strictly, a continuous cumulative distribution need not always have a density; singular continuous distributions are a mathematical exception. Most applied references use continuous distribution to mean the absolutely continuous, density-based case. That is the practical focus here.
Validate the model against relevant data, especially the tails used in decisions. Rounded measurements, censoring, truncation, and mixtures of different processes affect how a distribution should be fitted. A smooth curve alone is not evidence of a good fit.

Three practical examples

A service team models request duration to estimate the share exceeding a response-time target.
A manufacturer models variation in a measured dimension to estimate the proportion outside tolerance limits.
An AI team produces a conditional distribution for future demand so planners can compare uncertainty across different operating conditions.

Think of it this way

Think of probability as a fixed amount of paint spread along a strip. Some regions have thicker paint, but the amount covering a region depends on both thickness and width. A single infinitely thin cut contains no positive amount, even though it lies within the painted strip.

Worked example

Suppose a deliberately simplified model treats a duration X as uniformly distributed between two and six minutes. The density is 1/(6 - 2) = 0.25 per minute within that interval and zero outside it. Its total area is four minutes multiplied by 0.25 per minute, equal to one.
The probability of a duration between three and five minutes is (5 - 3) x 0.25 = 0.50. The probability of exceeding five minutes is (6 - 5) x 0.25 = 0.25. The 90th percentile is 2 + 0.90 x 4 = 5.6 minutes.
The probability of exactly four minutes is zero in this continuous model. However, if a recorder rounds to the nearest minute, a displayed value of four represents durations from 3.5 to 4.5 minutes, apart from irrelevant endpoint conventions. That interval has probability 1 x 0.25 = 0.25.
Expressed in seconds, the support is 120 to 360 and the density is 1/240 per second. The corresponding 180-to-300-second interval still has probability 120/240 = 0.50. The units change the density, not the underlying event probability.

Illustrative case study

Fictional case: Meadow Service uses an average task duration to promise completion times. Customers experience more late tasks than the average suggests, so analysts fit a continuous duration model and examine an upper quantile relevant to the promise.
They discover that one service type has a long right tail and that some unfinished tasks were excluded from the dataset. The team revises its data handling and evaluates separate distributions for materially different task types. Managers receive probabilities and model limitations alongside proposed targets. The exercise improves the questions asked about uncertainty; it does not prove a particular distribution remains suitable after staffing or workflow changes.

Common misconceptions and mistakes

Do not interpret a density of 0.25 as a 25% chance of one exact value. Do not assume every continuous distribution is symmetric, bell-shaped, or defined over all real numbers. A zero point probability does not make the realized observation impossible. Finally, recorded decimals do not establish continuity: a variable can take discrete permitted values even when those values contain decimal places.

Business use

Continuous distributions support capacity planning, simulation, quality analysis, and probabilistic forecasting. Select the decision-relevant quantity and evaluate the probabilities that actually affect costs or service commitments. Report units, distributional assumptions, parameter uncertainty, and evidence of fit. A useful business model should distinguish uncertainty in individual outcomes from uncertainty about the distribution's estimated parameters.

Check your understanding

Under the worked uniform model, what is P(X <= 3)?

Answer: One quarter.

Would changing minutes to seconds change that probability?

Answer: No, provided the event is converted consistently.

Does knowing the mean determine all interval probabilities?

Answer: No; distributions with the same mean can have very different spreads and tails.

Related terms

Learn next

Compare a uniform distribution and a skewed duration distribution using their cumulative probabilities and quantiles. Then explore how rounded observations and censored durations should enter the likelihood when estimating a continuous model.

Uniform distribution

Sources

Hossein Pishro-Nik, Introduction to Probability, Statistics, and Random Processes, Probability Density Function: https://www.probabilitycourse.com/chapter4/4_1_1_pdf.php KNIME, Data Science Glossary: https://www.knime.com/blog/data-science-glossary

Previous: Continuous optimization Next: Continuous random variable