Prior var

Computes variance of a prior distribution.

# S3 method for class 'prior'
var(x, ...)

Arguments

x

a prior

...

unused arguments

Value

a variance of an object of class 'prior'.

Details

The variance of a truncated normal prior is computed in closed form. Its terms cancel far in a tail and for narrow truncation intervals, so the variance is returned only when its estimated relative rounding error in double precision is at most 1e-6; otherwise the function stops with an error. For a one-sided truncation, the error is raised beyond about 33 prior standard deviations from the mean. The same applies to sd() and to spike-and-slab priors with such a slab. The variance can then be estimated from rng() draws of the prior, which sample the truncated distribution exactly, subject to Monte Carlo error. The variance of any other truncated prior is the integral of the squared deviation of its truncated quantile function from the mean over the probability interval (0, 1), non-negative by construction; it is refused with BayesTools_numerical_unavailable when the quantile function is unavailable or the integral's estimated error exceeds 1e-6 of the variance.

See also

prior()

Examples

# create a standard normal prior distribution
p1 <- prior(distribution = "normal", parameters = list(mean = 1, sd = 1))

# compute variance of the prior distribution
var(p1)
#> [1] 1