Pattern theory involves Pachinko fizzicks as Oberfell & Thompson realized long before anyone came up with PAM, which was predated by LDA [Latent Dirichlet Allocation], but still well after the penultimate werk done in Oklahoma by a couple of guys having a lot of fun earning their doctorates.

IOW, rather than get into a Bayesian inference or getting into the posterior of distribution, hard data remains vague, but one can recognize the footprint easily enough .. on a grease plate.