flexible_cfdr {fcfdr}

Perform Flexible cFDR

Description

Performs Flexible cFDR for continuous auxiliary covariates

Usage

flexible_cfdr(
  p,
  q,
  indep_index,
  res_p = 300,
  res_q = 500,
  nxbin = 1000,
  gridp = 50,
  splinecorr = TRUE,
  dist_thr = 0.5,
  locfdr_df = 10,
  plot = TRUE,
  maf = NULL,
  check_indep_cor = TRUE,
  enforce_p_q_cor = TRUE
)

Arguments

p

p-values for principal trait (vector of length n)

q

continuous auxiliary data values (vector of length n)

indep_index

indices of independent SNPs

res_p

number of grid points in x-direction (p) for KDE estimation

res_q

number of grid points in y-direction (q) for KDE estimation

nxbin

number of bins in x-direction (p) for hex-binning

gridp

number of data points required in a KDE grid point for left-censoring

splinecorr

logical value for whether spline correction should be implemented

dist_thr

distance threshold for spline correction

locfdr_df

df parameter in locfdr function

plot

logical value for whether to produce plots to assess KDE fit

maf

minor allele frequencies for SNPs to which p and q relate (optional and used to perform MAF matching)

check_indep_cor

check that sign of the correlation between p and q is the same in the independent subset as in the whole

enforce_p_q_cor

if p and q are negatively correlated, flip the sign on q values

Details

If maf is specified, then the independent SNPs will be down-sampled to match the minor allele frequency distribution.

Value

List of length two: (1) data.frame of p-values, q-values and v-values (2) data.frame of auxiliary data (q_low used for left censoring, how many data-points were left censored and/or spline corrected)

Examples

# this is a long running example
 
# In this example, we generate some p-values (representing GWAS p-values)
# and some arbitrary auxiliary data values (e.g. representing functional genomic data).
# We use the flexible_cfdr() function to generate v-values using default parameter values.

# generate p
set.seed(1)
n <- 1000
n1p <- 50 
zp <- c(rnorm(n1p, sd=5), rnorm(n-n1p, sd=1))
p <- 2*pnorm(-abs(zp))

# generate q
mixture_comp1 <- function(x) rnorm(x, mean = -0.5, sd = 0.5)
mixture_comp2 <- function(x) rnorm(x, mean = 2, sd = 1)
q <- c(mixture_comp1(n1p), mixture_comp2(n-n1p))

n_indep <- n

flexible_cfdr(p, q, indep_index = 1:n_indep)