HEM (version 1.44.0)

hem.eb.prior: Empirical Bayes (EB) Prior Specification

Description

Estimates experimental and biological variances by LPE and resampling

Usage

hem.eb.prior(dat,  n.layer, design,  method.var.e="neb", method.var.b="peb", method.var.t="neb",  rep=TRUE, baseline.var="LPE", p.remove=0, max.chip=4, q=0.01, B=25, n.digits=10, print.message.on.screen=TRUE)

Arguments

dat

data

n.layer

number of layers

design

design matrix

method.var.e

prior specification method for experimental variance; "peb"=parametric EB prior specification, "neb"=nonparametric EB prior specification

method.var.b

prior specification method for biological variance; "peb"=parametric EB prior specification

method.var.t

prior specification method for total variance; "peb"=parametric EB prior specification, "neb"=nonparametric EB prior specification

rep

no replication if FALSE

baseline.var

baseline variance estimation method: LPE for replicated data and BLPE, PSE, or ASE for unreplicated data

p.remove

percent of removed rank-variance genes for BLPE

max.chip

maximum number of chips to estimate errors

q

quantile for paritioning genes based on expression levels

B

number of iterations for resampling

n.digits

number of digits

print.message.on.screen

if TRUE, process status is shown on screen.

Value

var.b

prior estimate matrix for biological variances (n.layer=2)

var.e

prior estimate matrix for experiemtnal variances (n.layer=2)

var.t

prior estimate matrix for total variances (n.layer=1)

See Also

hem, hem.fdr

Examples

#Example 1: Two-layer HEM with EB prior specification

data(pbrain)

##construct a design matrix
cond <- c(1,1,1,1,1,1,2,2,2,2,2,2)
ind  <- c(1,1,2,2,3,3,1,1,2,2,3,3)
rep  <- c(1,2,1,2,1,2,1,2,1,2,1,2)
design <- data.frame(cond,ind,rep)

##normalization
pbrain.nor <- hem.preproc(pbrain[,2:13])

##take a subset for a testing purpose;
##use all genes for a practical purpose
pbrain.nor <- pbrain.nor[1:1000,]

##estimate hyperparameters of variances by LPE
#pbrain.eb  <- hem.eb.prior(pbrain.nor, n.layer=2,  design=design,
#                           method.var.e="neb", method.var.b="peb")     

#fit HEM with two layers of error
#using the small numbers of burn-ins and MCMC samples for a testing purpose;
#but increase the numbers for a practical purpose 
#pbrain.hem <- hem(pbrain.nor, n.layer=2,  design=design,burn.ins=10, n.samples=30, 
#              method.var.e="neb", method.var.b="peb", 
#              var.e=pbrain.eb$var.e, var.b=pbrain.eb$var.b)


#Example 2: One-layer HEM with EB prior specification

data(mubcp)

##construct a design matrix
cond <- c(rep(1,6),rep(2,5),rep(3,5),rep(4,5),rep(5,5))
ind  <- c(1:6,rep((1:5),4))
design <- data.frame(cond,ind)

##normalization
mubcp.nor <- hem.preproc(mubcp)

##take a subset for a testing purpose;
##use all genes for a practical purpose
mubcp.nor <- mubcp.nor[1:1000,] 

##estimate hyperparameters of variances by LPE
#mubcp.eb  <- hem.eb.prior(mubcp.nor, n.layer=1, design=design,
#             method.var.t="neb")                                

#fit HEM with two layers of error
#using the small numbers of burn-ins and MCMC samples for a testing purpose;
#but increase the numbers for a practical purpose 
#mubcp.hem <- hem(mubcp.nor, n.layer=1, design=design,  burn.ins=10, n.samples=30, 
#             method.var.t="neb", var.t=mubcp.eb$var.t)