Double Wilson Prior Calculations in careless
This page briefly describes the so-called “Double Wilson” distributions available for use in careless when using one dataset as a reference for another.
We’d like to be able to make priors in which certain sets of structure factors are conditional on others. The conditional distributions for structure factors are the Rice and Woolfson (Folded Normal) distributions for acentrics and centrics respectively.
Wilson’s priors
Joint priors
For two structure factors, \(G_h\) and \(F_h\), let
denote reparameterized samples. The joint probability factorizes as, $\(P(F_h, G_h) = P(G_h | F_h) P(F_h).\)\( \)F_h\( follows Wilson's prior, \)\(P(F_h) = \mathcal{W}ilson(\epsilon_h, \Sigma),\)$
Acentric joint priors
The joint probability of the samples is
Where the conditional is a Rice distribution, $\( Rice(x | \nu, \sigma) = \frac{x}{\sigma^2} \exp\left(\frac{-(x^2+\nu^2)}{2\sigma^2}\right)I_0\left(\frac{x\nu}{\sigma^2}\right) \)$
Centric joint priors
The joint probability of the samples is
Where the conditional is a folded normal distribution, $\( FoldedNormal(x | \mu, \sigma) = \frac{1}{\sigma\sqrt{2\pi}}\exp\left(-\frac{(x-\mu)^2}{2\sigma^2}\right) + \frac{1}{\sigma\sqrt{2\pi}}\exp\left(-\frac{(x+\mu)^2}{2\sigma^2}\right) \)$