MRCIP: MRCIP: A robust Mendelian randomization method accounting for...

In siqixu/MRCIP: MRCIP: A robust Mendelian randomization method accounting for correlated and idiosyncratic pleiotropy

Description

MRCIP is a Mendelian randomization approach for the inference of the causal effect of an exposure variable on an outcome of interest based on summary-level GWAS data, which accounts for the correlated pleiotropy and idiosyncratic pleiotropy.

Usage

MRCIP(data,beta=0,mu_gamma=NA,s_gamma=NA,s_alpha=NA,rho=0,h=NA,h_start=0.01,h_step=0.01,tol_dw=1e-4,tol=1e-8,n_iter=1000,MVNtest="FALSE")

Arguments

data	A matrix or data frame consists of four columns: the 1st (2nd) column contains the estimated genetic effects on the outcome (exposure); the 3rd (4th) column contains the estimated standard errors of the estimated genetic effects on the outcome (exposure).
beta	The initial value of the causal effect in the PRW-EM algorithm. The default value is zero.
mu_gamma	The initial value for the mean of the genetic effects on the exposure in the PRW-EM algorithm. If it is missing, the sample mean of the estimated genetic effects on the exposure will be used.
s_gamma	The initial value for the standard error of the genetic effects on the exposure in the PRW-EM algorithm. If it is missing, it will be calculated based on the sample variance of the estimated genetic effects on the exposure.
s_alpha	The initial value for the standard error of the pleiotropy effects on the outcome in the PRW-EM algorithm. If it is missing, it will be calculated based on the sample variance of the estimated genetic effects on the outcome.
rho	The initial value of the correlated pleiotropy index (CPI) in the PRW-EM algorithm. The default value is zero.
h	The bandwidth used in the kernel estimation for the Pearson residuals. If it is missing, the value "sel_h" provided in the plot_h function will be used, which detects the possible abrupt change in the empirical mean down-weighting level over a grid of h values.
h_start	The start value in the grid search for the bandwidth used in the kernel estimation. The default value is 0.01. It will be ignored if h is provided.
h_step	The step size in the grid search for the bandwidth used in the kernel estimation. The default value is 0.01. It will be ignored if h is provided.
tol_dw	The grid search for the bandwidth stops when the difference in the mean downweighting level between two consecutive steps is less than tol_dw. It will be ignored if h is provided.
tol	The tolerance used in the PRW-EM algorithm. The PRW-EM algorithm stops when the difference in the weighted log-likelihood function between two consecutive iterations is less than tol. The default value is 1e-8.
n_iter	The maximum number of iterations in the PRW-EM algorithm. The default value is 1000.
MVNtest	If MVNtest="TRUE", then the chi-square QQ plot and the Mardia's test are provided for the multivariate normality test of the standardized residuals.

Value

beta	The estimated causal effect of the exposure on the outcome, the corresponding p value, the lower limit and upper limit of the 95 percent confidence interval
rho	The estimated correlated pleiotropy index (CPI) as well as the corresponding p value
IVs with weights less than 0.5	The indices and the weights of the genetic variants that have the weights less than 0.5
weights	The weights of the genetic variants calculated by the Pearson residuals
Normality test for the standardized residuals	The Mardia’s test for the multivariate skewness and kurtosis of the standardized residuals
standardized residuals	The standardized residuals of the MRCIP model

Examples

  out = MRCIP(data=example)
  out$beta # the estimation and inference for the causal effect
  out$rho # the estimation and inference for correlated pleiotropy index (CPI)

siqixu/MRCIP documentation built on Jan. 28, 2021, 9:42 a.m.