Priors

`MultipleIndependent`

`max_log_pdf_per_axis(n_grid=1024)`

Per-axis upper bound on logpdf over the prior support.

Used by least-squares MAP fits to shift the prior log-density into a non-negative range, so that the prior contribution can be encoded as extra residuals whose sum of squares equals -2 * log_prior + const. The grid combines an inverse-CDF sweep (denser where the prior is concentrated) with the support endpoints, which suffices for the priors in this package (uniform, loguniform, truncated norm).

`build_prior(define_prior)`

Build the joint prior from a list of XSPEC prior specifications.

For a single-model fit, each item must be a 3-tuple:

(component_name, parameter_name, distribution)

For a multi-model fit, each item must be a 4-tuple:

(model_name, component_name, parameter_name, distribution)

distribution can be and must behave like a SciPy continuous distribution and provide at least support(), ppf(), cdf(), rvs(), and logpdf(). Component and parameter names must match the names exposed by xspec. Repeated components can be addressed with suffixed names such as powerlaw_3.

Parameters:

Name	Type	Description	Default
`define_prior`		Sequence describing one prior per free XSPEC parameter.	required

Returns:

Type	Description
	`(prior, parameters_index, model_index)`
	`prior` is a `MultipleIndependent`
	distribution over all unfrozen, unlinked fitted parameters. Its
	per-parameter support bounds are available on `prior.bounds`.

`loguniform(low, high)`

Create a log-uniform prior on a positive interval.

Parameters:

Name	Type	Description	Default
`low`		Strictly positive lower bound.	required
`high`		Upper bound.	required

Returns:

Type	Description
	A SciPy `loguniform` distribution defined on `[low, high]`.

`norm(loc, scale, low, high)`

Create a truncated Gaussian prior.

Parameters:

Name	Description	Default
`loc`	Mean of the underlying Gaussian distribution.	required
`scale`	Standard deviation of the underlying Gaussian distribution.	required
`low`	Lower truncation bound, expressed in SciPy `truncnorm` coordinates.	required
`high`	Upper truncation bound, expressed in SciPy `truncnorm` coordinates.	required

Returns:

Type	Description
	A SciPy `truncnorm` distribution with the requested location,
	scale, and truncation limits.

`uniform(low, high)`

Create a uniform prior on a finite interval.

Parameters:

Name	Type	Description	Default
`low`		Lower bound of the interval.	required
`high`		Upper bound of the interval.	required

Returns:

Type	Description
	A SciPy `uniform` distribution defined on `[low, high]`.