psi variable not used in compute_mallows?

[lorenzo-zuccato]

Hi, I have noticed that the psi parameter is not used anywhere in the function compute_mallows. I think it should be passed to the run_mcmc function, otherwise it will always use 10 as fixed value. Is there a reason for this?

[osorensen]

Thanks @lorenzo-zuccato, that's a very good catch.

[osorensen]

Here is a test I've made after properly propagating the psi argument to run_mcmc. It seems to me correct, in light of the animation on Wikipedia, that with lower psi we get closer to hard clustering, and with increasing psi we get soft clustering.

library(BayesMallows)
library(ggplot2)
library(patchwork)
set.seed(123)
mixture_model1 <- compute_mallows(
  rankings = sushi_rankings[1:100, ], n_clusters = 3,
  include_wcd = TRUE, save_clus = TRUE, nmc = 100, psi = 100
)

set.seed(123)
mixture_model2 <- compute_mallows(
  rankings = sushi_rankings[1:100, ], n_clusters = 3,
  include_wcd = TRUE, save_clus = TRUE, nmc = 100, psi = .1
)

(assess_convergence(mixture_model1, parameter = "cluster_probs") +
  ggtitle("psi = 100")) +
  (assess_convergence(mixture_model2, parameter = "cluster_probs") +
     ggtitle("psi = 0.1"))

^{Created on 2023-10-03 with reprex v2.0.2}

[osorensen]

Thanks again @lorenzo-zuccato, this is fixed now, and will be on CRAN in a few days, hopefully.