Prior probability
Also known as: Prior
What you believed about a hypothesis before seeing the new data.
What it means
A prior probability is the probability assigned to a hypothesis or parameter before incorporating the current data, representing background knowledge or assumptions. In Bayesian inference it is combined with the likelihood to yield the posterior, so the prior's influence is strongest when data are scarce and fades as evidence accumulates. Priors can be informative, drawing on previous studies or theory, or deliberately weak and 'uninformative' to let the data dominate. Their subjectivity is the most debated feature of Bayesian methods, addressed in practice through sensitivity analysis across plausible priors.
Examples
Before a clinical trial, a researcher's prior might place most plausibility on small or null drug effects, reflecting that most candidates fail.
Hearing hoofbeats, expect horses, not zebras: the sheer commonness of horses is your prior, and one ambiguous sound is nowhere near enough evidence to overturn it.
A friend swears she saw a film star in the corner shop. Your prior says unlikely, so a glimpse won't do — but a clear photo would move you.
First described in Bayesian tradition; Bayes (1763), Laplace.