Priorval

Priorval, short for prior value, is a term sometimes used in Bayesian statistics and probabilistic programming to denote the value assigned to a parameter before observing data. Unlike the prior distribution itself, which is a formal probability model, priorval is typically a concrete numerical or vector value used to seed priors or initialization processes, and its exact meaning depends on context or software.

In many libraries, priorval serves as the starting point for optimization or as the mean (or mode) around which a prior distribution is centered. For example, a parameter p with a bounded interval [0,1] might be assigned priorval = 0.5, with a beta prior described around that center, or in a normal prior with mean equal to priorval.

It is important to distinguish priorval from the posterior or from the hyperparameters of the prior. The priorval itself is not a probability, but a value used to construct, constrain, or initialize the prior. Its use and interpretation vary across frameworks; some codebases treat priorval as a hard initial guess, others use it to set the prior's central tendency.

Within documentation, priorval is sometimes represented as a scalar or vector aligning with the parameter's dimensionality. It often reflects domain knowledge or empirical estimates before data are observed. See also prior, prior distribution, Bayesian inference.