mbglm: Fitting Bayesian GLMs for Many Responses

In nyiuab/BhGLM: Bayesian hierarchical GLMs and survival models, with applications to Genomics and Epidemiology

Description

mbglm fits Bayesian GLMs separately for many responses, and summary.mbglm summarizes the fitted models.

Usage

  
mbglm(y, formula, data, family=NegBin(), prior=Student(0,1),
      min.p=0, verbose=TRUE)
      
summary.mbglm(object)

Arguments

y	a matrix of responses; each column is a response and rows are samples.
formula	a one-sided formula of the form ~ x (i.e., the respose is omitted); the right side of ~ is the same as in bglm.
data, family, prior, verbose	These arguments are the same as in bglm.
min.p	a value in [0, 1). The responses with the proportion of non-zero values > min.p are analyzed.
object	an object from mbglm.

Details

This function analyzes the responses in y by repeated calls to bglm.

Value

The function mbglm returns a list including fit, responses, variables, and call:

fit	fitted models for all the responses;
responses	names of all the responses;
variables	names of all covariates;

The function summary.mbglm returns a data frame consisting of responses, variables, estimates, standard deviations, p-vlaues and adjusted p-vlaues (using FDR method) for all coefficients.

Author(s)

Nengjun Yi, [e-mail]

See Also

bglm

Examples

library(BhGLM)
library(NBZIMM)

# load data in NBZIMM
data(Romero)
names(Romero)

otu = Romero$OTU; dim(otu)
sam = Romero$SampleData; dim(sam)
colnames(sam)

Days = sam$GA_Days; Days = scale(Days)
Age = sam$Age; Age = scale(Age)
Race = sam$Race
preg = sam$pregnant; table(preg)

# calculate size factor using the cumulative sum scaling (CSS) method in metagenomeSeq
library(metagenomeSeq)
obj = newMRexperiment(counts=t(otu), phenoData=NULL, featureData=NULL,
                      libSize=NULL, normFactors=NULL)
obj = cumNorm(obj, p=cumNormStat(obj)) #set p=1: generate total reads
s = normFactors(obj)
s = unlist(s)

# analyze taxa with nonzero proportion > min.p
f = mbglm(y=otu, formula=~ Days + Age + Race + preg + offset(log(s)), 
          family=NegBin(), prior=Student(0, 1), min.p=0.1)

out = summary.mbglm(f)
out = out[out[,2]=="preg", ]
coefs = out[,3]; names(coefs) = out[,1]
sds = out[,4]
padj = out[,6]
plot.bh(coefs=coefs, sds=sds, pvalues=padj, threshold=0.001, gap=1000)

out = summary.mbglm(f)
df = out[,c(1:4, 6)]; colnames(df)[5] = "pvalue" 
g = NBZIMM::heat.p(df=df, p.breaks = c(0.001, 0.01, 0.05), 
           colors = c("black", "darkgrey", "grey", "lightgrey"),
           zigzag=c(T,F), abbrv=c(T,F), margin=c(2.5,0.5), y.size=8,
           legend=T)

nyiuab/BhGLM documentation built on Jan. 9, 2022, 3:31 p.m.