qgg (version 1.1.6)

ldsc: LD score regression

Description

The ldsc function is used for LDSC analysis

Usage

ldsc(
  Glist = NULL,
  ldscores = NULL,
  sets = NULL,
  method = "regression",
  residual = FALSE,
  z = NULL,
  b = NULL,
  seb = NULL,
  af = NULL,
  stat = NULL,
  tol = 1e-08,
  n = NULL,
  intercept = TRUE,
  what = "h2",
  maxZ2 = NULL,
  SE.h2 = FALSE,
  SE.rg = FALSE,
  blk = 200
)

Value

Returns a matrix of heritability estimates when what="h2", and if SE.h2=TRUE standard errors (SE) and significance levels (P) are returned. If what="rg" an n-by-n matrix of correlations is returned where the diagonal elements being h2 estimates. If SE.rg=TRUE a list is returned with n-by-n matrices of genetic correlations, estimated standard errors and significance levels.

Arguments

Glist

list of information about genotype matrix stored on disk

ldscores

vector of LD scores (optional as LD scores are stored within Glist)

sets

Optional list specifying sets of SNPs for mapping.

method

the regression method to use, options include "regression", "bayesC", "bayesR".

residual

logical if TRUE then add a residual that capture the h2 not explained by the sets

z

matrix of z statistics for n traits

b

matrix of marker effects for n traits if z matrix not is given

seb

matrix of standard errors of marker effects for n traits if z matrix not is given

af

vector of allele frequencies

stat

dataframe with marker summary statistics

tol

smallest value for h2

n

vector of sample sizes for the traits (element i corresponds to column vector i in z matrix)

intercept

logical if TRUE the LD score regression includes intercept

what

either computation of heritability (what="h2") or genetic correlation between traits (what="rg")

maxZ2

maximum value for squared value of z-statistics

SE.h2

logical if TRUE standard errors and significance for the heritability estimates are computed using a block jackknife approach

SE.rg

logical if TRUE standard errors and significance for the genetic correlations are computed using a block jackknife approach

blk

numeric size of the blocks used in the jackknife estimation of standard error (default = 200)

Author

Peter Soerensen

Palle Duun Rohde

Examples

# Plink bed/bim/fam files
 #bedfiles <- system.file("extdata", paste0("sample_chr",1:2,".bed"), package = "qgg")
 #bimfiles <- system.file("extdata", paste0("sample_chr",1:2,".bim"), package = "qgg")
 #famfiles <- system.file("extdata", paste0("sample_chr",1:2,".fam"), package = "qgg")
 #
 ## Summarize bed/bim/fam files
 #Glist <- gprep(study="Example", bedfiles=bedfiles, bimfiles=bimfiles, famfiles=famfiles)

 #
 ## Filter rsids based on MAF, missingness, HWE
 #rsids <-  gfilter(Glist = Glist, excludeMAF=0.05, excludeMISS=0.05, excludeHWE=1e-12) 
 #
 ## Compute sparse LD (msize=size of LD window)
 ##ldfiles <- system.file("extdata", paste0("sample_chr",1:2,".ld"), package = "qgg")
 ##Glist <- gprep(Glist, task="sparseld", msize=200, rsids=rsids, ldfiles=ldfiles, overwrite=TRUE)
 #
 #
 ##Simulate data
 #W1 <- getG(Glist, chr=1, scale=TRUE)
 #W2 <- getG(Glist, chr=2, scale=TRUE)

 #W <- cbind(W1,W2)
 #causal <- sample(1:ncol(W),5)

 #b1 <- rnorm(length(causal))
 #b2 <- rnorm(length(causal))
 #y1 <- W[, causal]%*%b1 + rnorm(nrow(W))
 #y2 <- W[, causal]%*%b2 + rnorm(nrow(W))

 #data1 <- data.frame(y = y1, mu = 1)
 #data2 <- data.frame(y = y2, mu = 1)
 #X1 <- model.matrix(y ~ 0 + mu, data = data1)
 #X2 <- model.matrix(y ~ 0 + mu, data = data2)

 ## Linear model analyses and single marker association test
 #maLM1 <- lma(y=y1, X=X1,W = W)
 #maLM2 <- lma(y=y2,X=X2,W = W)
 #
 ## Compute heritability and genetic correlations for trait 1 and 2
 #z1 <- maLM1[,"stat"]
 #z2 <- maLM2[,"stat"]

 #z <- cbind(z1=z1,z2=z2)

 #h2 <- ldsc(Glist, z=z, n=c(500,500), what="h2")
 #rg <- ldsc(Glist, z=z, n=c(500,500), what="rg")