get_inits_overdispersed_four_chains: Helper function to define overdispersed initial values
In snsf-data/ERforResearch: Expected Rank for Research Evaluation

get_inits_overdispersed_four_chains

R Documentation

Helper function to define overdispersed initial values

Description

Note that this function can only be applied when four chains are used, otherwise the user has to provide the initial values themselves. It is however recommended to use four chain. The helper function will also use the four different random generators provided by rjags.

Usage

get_inits_overdispersed_four_chains(
  merging_panels = FALSE,
  ordinal_scale = FALSE,
  point_scale = NULL,
  heterogeneous_residuals = FALSE,
  n_proposals,
  n_assessors,
  n_panels = NULL,
  grades = NULL,
  seed
)

Arguments

`merging_panels`	should the sections be merged?, default is FALSE, e.g. a ranking is done for each section/panel separately.
`ordinal_scale`	are the evaluation scores on an ordinal scale?, default is FALSE.
`point_scale`	if on an ordinal scale, what is the number of points on the scale?, default is NULL
`heterogeneous_residuals`	dummy variable informing us on whether or not the residuals should be heterogeneous, default = FALSE.
`n_proposals`	number of applications / proposals considered.
`n_assessors`	number of voters / panel members.
`n_panels`	number of sections / panels, if merged, default is NULL.
`grades`	the grades given to the proposals (the whole n_proposals*n_assessors long vector)
`seed`	seed used for the sampling

Details

It is optimal to use four chains. This also allows the sampling to use the four different samplers implemented in rjags (e.g. Wichmann-Hill, Marsaglia-Multicarry, Super-Duper, Mersenne-Twister)