term.reparam_noncentered()

term.reparam_noncentered()

Turns this term into noncentered form, which means the prior for the coefficient will be turned from coef ~ N(0, scale^2 * inv(penalty)) into latent_coef ~ N(0, inv(penalty)); coef = scale * latent_coef. This can sometimes be helpful when sampling with the No-U-Turn Sampler.

Return type:

Self