postdrawspcd: Draws samples from the posterior distribution for paired...

In oseipep/pcdbayeslm: Bayesian analysis of paired comparison data

Draws samples from the posterior distribution for paired comparison data

Description

This function performs a new posterior ranking methodology developed in the paper by Prince P. Osei and Ori David (2021) to account for the uncertainty in the ranking of items.

Usage

postdrawspcd(
  N,
  noitems,
  nocompars,
  scores,
  vars = 1,
  xmu = zeros(noitems, 1),
  xvar = vars * diag(noitems),
  a0 = 2,
  b0 = 1,
  Edges = c(1, 2, 1, 3, 2, 3),
  data = NULL,
  datatype = "simulated",
  prior = "conju"
)

Arguments

N	Number of posterior samples.
noitems	Number of items in the comparison graph
nocompars	The number of comparisons among noitems.
scores	The true scores or merits of noitems
vars	The constant variance of scores default is 1.
xmu	The normal prior mean of noitems default is a vector of zeros of length noitems.
xvar	The prior covariance matrix of noitems default is vars times the identity matrix dimension noitems.
a0	The prior shape of inverse gamma distribution default is 2
b0	The prior scale of inverse gamma distribution default is 1
Edges	The edge set of noitems in the graph default is complete graph of three noitems.
data	The data containing the comparison outcomes; defalut is NULL.
datatype	The type of data to analyze: "simulated" (the default) for simulated data, "real" for a given real dataset.
prior	The type of prior: "conju" (default) for conjugate prior, "semi-conju" for semi-conjugate prior, "flat" for flat prior, and "ref" for reference prior

Value

A posterior rankings of posterior draws from cpcbayeslm

Author(s)

Prince P. Osei and Ori Davidov

Examples

##  A graph of three items
##  K <- 3 # number of items
## edges <- c(1,2,1,3,2,3) # edge set
## compars <- rep(3,3) # pairwise comparisons
##  Tscores <- 1:3-mean(1:3) # The true score sum to zero
##  N <- 100 # number of samples to draw
## postdrawspcd(N,K,compars,Tscores,Edges=edges)

oseipep/pcdbayeslm documentation built on Aug. 6, 2023, 8:45 a.m.