vb_fit_family: Fit Variational Bayes model for family data

In yhai943/FBLMM: Bayesian linear mixed model with multiple random effects

Fit Variational Bayes model for family data

Description

Take in genotype, phenotype and analyses the target sequence by using BLMM.

Usage

vb_fit_family(
  y,
  genotype0,
  max_iter = 10000,
  weight.type = NULL,
  maf.filter.snp = 0.01,
  epsilon_conv = 1e-04,
  Bsigma2beta = 1,
  theta_beta = 0.1,
  theta_u = 0.1,
  verbose = TRUE,
  kernel = "Lin",
  kin.mat = kin.mat,
  dummy.fam = dummy.fam
)

Arguments

y	a phenotype vector of length n
genotype0	a list of genotype matrices of the target sequence
max_iter	maximum number of iteration
weight.type	type of weight function
maf.filter.snp	a filtering threshold of minor allele frequency for the isolated predictors
epsilon_conv	a numeric value giving the interval endpoint
Bsigma2beta	a numeric value for sigma beta
theta_beta	probability of causal variants
theta_u	probability of causal regions
verbose	informative iteration messages
kernel	kernel type for covariance matrix
kin.mat	kinship covariance matrix
dummy.fam	family indicator matrix

Details

A hybrid model that includes a sparsity regression model and a LMM with multiple random effects. The sparsity regression model part is designed to capture the strong predictive effects from isolated predictors, whereas the LMM part is to capture the effects from a group of predictors located in nearby genetic regions.

Examples

datafile = system.file("extdata",package = "BLMM")
data <- read_file(gene_file = paste(datafile,"snp.bed",sep="/"),y_file = paste(datafile,"y.txt",sep="/"),kin.mat=T)
y = data$y
gene = data$genematrix
# Extract gene intervals from BED files
genotype0 <- list(as.matrix(gene[,1:200]), as.matrix(gene[,300:450]))
dummy.fam = data$dummy.fam
kin.mat = data$kin.mat
# choose model type: 1. "uw" for BLMM-UW; 2. "beta" for BLMM-BETA; 3. "wss" for BLMM-WSS;
fit <- vb_fit_family(y = y, genotype0 = genotype0,dummy.fam = dummy.fam,kin.mat=kin.mat)

yhai943/FBLMM documentation built on March 19, 2022, 8:21 a.m.