pymc.WishartBartlett#

pymc.WishartBartlett(name, S, nu, is_cholesky=False, return_cholesky=False, initval=None)[source]#

Bartlett-decomposed Wishart prior. Deprecated: use Wishart directly.

This used to be the only MCMC-usable Wishart in PyMC, since the legacy Wishart had no unconstraining transform. The new Wishart is parameterized through its Cholesky factor with a default CholeskyCovTransform, so this helper is no longer needed and is a thin shim around it for backward compatibility.

Parameters:

Sndarray: (p, p) positive-definite scale matrix, or its lower-triangular Cholesky factor when is_cholesky=True.
nutensor_like of int: Degrees of freedom, > dim(S) - 1.
is_choleskybool, default False: If True, S is interpreted as the Cholesky factor of the scale matrix (mapped to Wishart’s scale_chol argument).
return_choleskybool, default False: If True, return the Cholesky factor of the Wishart draw rather than the SPD matrix itself.
initvalndarray, optional: Forwarded to Wishart. Note that initval semantics changed: the new Wishart expects an SPD matrix value, not the legacy Bartlett component initvals.