The Greiner-Quinn method involves a Gibbs sampler that contains a Metropolis Hastings step within it for sampling the variable omega (a transformation of the voting preferences). Within this MH step he proposal distribution for omega is a t_4 distribution with variance scaled by a term gamma. Right now, the user can specify a global gamma for all precincts, but in the R package RxCEcolinf, there is a different gamma in each precinct, and it is chosen (during a tuning phase) with the objective of acheiving acceptance probabilities between .2 and .5. See docs:https://cran.r-project.org/web/packages/RxCEcolInf/RxCEcolInf.pdf (p. 24). It would be good to add similar tuning of gamma in our implementation.
The Greiner-Quinn method involves a Gibbs sampler that contains a Metropolis Hastings step within it for sampling the variable omega (a transformation of the voting preferences). Within this MH step he proposal distribution for omega is a t_4 distribution with variance scaled by a term gamma. Right now, the user can specify a global gamma for all precincts, but in the R package RxCEcolinf, there is a different gamma in each precinct, and it is chosen (during a tuning phase) with the objective of acheiving acceptance probabilities between .2 and .5. See docs:https://cran.r-project.org/web/packages/RxCEcolInf/RxCEcolInf.pdf (p. 24). It would be good to add similar tuning of gamma in our implementation.