Transformations to the Prior Density Scale

prior_transform_exp() describes exp(scale * x) for an unconstrained model parameter x. prior_transform_square() describes x^2 on the non-negative branch only. Both map to a non-negative density scale, with strictly positive values in the interior of the parameter domain.

Usage

prior_transform_exp(scale = 1)

prior_transform_square()

Arguments

scale

(number) finite positive multiplier inside the exponential. Use 2 for a prior on the variance when x is a log standard deviation.

Value

A PriorTransform object.

Details

Supply these objects as the transform argument to a scalar distribution helper such as prior_normal() or to prior_transform().

The intended density is p(g(x)) * abs(g'(x)), including the Jacobian. The square transformation requires a non-negative model parameter so that it is one-to-one. It must not silently fold an unconstrained parameter.

Examples

prior_transform_exp()
#> An object of class "PriorTransform"
#> Slot "kind":
#> [1] "exp"
#> 
#> Slot "scale":
#> [1] 1
#> 
prior_transform_exp(scale = 2)
#> An object of class "PriorTransform"
#> Slot "kind":
#> [1] "exp"
#> 
#> Slot "scale":
#> [1] 2
#> 
prior_transform_square()
#> An object of class "PriorTransform"
#> Slot "kind":
#> [1] "square"
#> 
#> Slot "scale":
#> [1] 1
#> 
prior_normal(60, 60, transform = prior_transform_exp())
#> 
#> Prior Object:
#>    normal(mu = 60, sigma = 60) T[0, ] on exp(x)
#>