BiG (version 0.1.0)

BiG_diffuse: BiG with diffuse Inverse Gamma/Uniform prior

Description

BiG implemented with diffuse Inverse Gamma prior or diffuse Uniform prior for the variance/standard deviation parameters.

Usage

BiG_diffuse(r, n_T, n_p1, M = 20000, burnin, prior, ds = 1, dp = 1,
  W = init_W(r), sigma_p10 = 0.5, sigma_p20 = 0.5, mu0 = numeric(G),
  kappa10 = numeric(G), kappa20 = numeric(G), sigma_s0 = rep(1, S),
  a = 0.0202, b = 98.5025)

Arguments

r

G*S matrix that contains the ranked lists to be aggregated, where G is the total number of items (genes) and S is the total number of ranked lists (studies). NA indicates non-inclusion of item. Note the matrix needs to be arranged such that all lists that belong to the same platform are next to each other, i.e. the first n_p1 columns are lists from platform 1.

n_T

vector of length S that contains number of top ranked items in each study.

n_p1

number of studies belong to platform 1.

M

number of MCMC iterations.

burnin

number of burn-in iterations.

prior

either "IG" or "uniform"

ds, dp

hyperparameter for the prior distributions of variance parameters for study bias and platform bias respectively. Used only when prior="IG".

W

G*S matrix that contains initial values for W. Each element of W is the local importance of the corresponding item in the corresponding study, i.e. the latent variable that determines the observed rank.

sigma_p10, sigma_p20

initial values for the variance of the platform bias for platform 1 and platform 2 respectively.

mu0

vector of length G that contains initial values for mu. Each element of mu is global importance of the corresponding item, i.e. the latent variable that determines the true rank.

kappa10, kappa20

vectors of length G that contain initial values for kappa1 and kappa2. Each element of kappa1 and kappa2 is the platform bias for the corresponding item in platform 1 and 2 respectively

sigma_s0

vector of length S that contains initial values for the variances of the study bias.

a, b

hyperparameters for the prior distributions of standard deviation parameters. Used only when prior="uniform".

Examples

# NOT RUN {
set.seed(1234)
sim = sim_lvm(G=25, S=6, n_p1=3, rho=runif(6,min=0.3,max=0.9), p_p1=0.6, p_p2=0.8, 
      lambda=runif(6,min=0.6,max=0.8), n_T=sample(c(5,10,15),6,replace=TRUE))
rank(-BiG_diffuse(r=sim$r, n_T=sim$n_T, n_p1=3, M=100, burnin=50, prior="IG"))
#rank(-BiG_diffuse(r=sim$r, n_T=sim$n_T, n_p1=3, M=100, burnin=50, prior="uniform"))
# }