Continuous Distribution Atlas
Which continuous distribution matches the story behind the data?
How this lesson unfolds.
Begin with How do physical constraints and generating mechanisms dictate which continuous distribution correctly models real data? Use the picture, intuition and explanation together before moving on.
How do physical constraints and generating mechanisms dictate which continuous distribution correctly models real data?
Every distribution is an equation born of a physical physics constraint: waiting for an unpredictable bus generates an exponential curve, measuring the sum of tiny independent measurement errors shapes a Gaussian bell, and choosing a random angle on a spinner flattens into a uniform slab. Identify the generative mechanism of uncertainty, and the correct mathematical distribution chooses itself.
- $$f_X(x) \ge 0, \quad \int_{\text{supp}(X)} f_X(x)\,dx = 1$$
- Learning goalSelect continuous probability models by matching empirical data to generative support constraints.
- Helpful foundationsContinuous random variables · Probability density functions · Support and intervals · Foundations of the Theory of Probability (Andrey Kolmogorov, 1933) · An Essay towards solving a Problem in the Doctrine of Chances (Thomas Bayes, 1763)
- Common trapDistribution selection requires verifying support bounds and physical mechanisms, not superficial curve-fitting.
- Next connectionExplore uniform and exponential models as foundational baselines for continuous probability.
More context & mechanics
Continuous distributions are not arbitrary curves—each arises from a strict generating process with distinct mathematical support:
- Uniform on [a, b]: Total ignorance; constant density f(x)=1/(b-a).
- Exponential on [0, ∞): Memoryless waiting time for the next Poisson event with rate λ.
- Normal on (-∞, ∞): Sum of many small, independent perturbations by the Central Limit Theorem.
- Gamma on [0, ∞): Time elapsed until k independent Poisson events occur.
- Beta on [0, 1]: Prior and posterior uncertainty over an unknown proportion or probability.
- Log-Normal on (0, ∞): Multiplicative compounding of independent positive shocks (X = e^Y).
In cloud infrastructure engineering, site reliability engineers (SREs) model microservice API latencies using Log-Normal distributions because serialization, network transit, and queuing delays multiply together. Using a Gaussian model would assign physically impossible nonzero probabilities to negative execution times.
In 1895, English mathematician Karl Pearson noticed that natural biological data rarely fit Gauss's symmetric bell curve. Pearson developed a systematic 4-parameter differential equation system that unified the Beta, Gamma, and Pareto families into a single structural atlas classified by skewness and kurtosis.
Remember the 'Generative Story Grid': Flat bounded = Uniform; One wait = Exponential; k waits = Gamma; Additive sum = Normal; Multiplicative growth = Log-Normal; Probability of a rate = Beta.