Primo_tstat: Estimate posterior probabilities of association patterns,...

In kjgleason/primo: Package in R for Integrative Multi-Omics association analysis

Description

For each observation (e.g. SNP), estimates the posterior probability of each association pattern. Utilizes parallel computing, when available.

Usage

Primo_tstat(betas, sds, dfs, alt_props, mafs = NULL, N = NULL,
  Gamma = NULL, tol = 0.001, par_size = 1)

Arguments

betas	matrix of coefficient estimates.
sds	matrix of standard errors (for coefficient estimates).
dfs	vector or matrix of degrees of freedom.
alt_props	vector of the proportions of test statistics from alternative densities.
mafs	vector or matrix of minor allele frequencies (MAFs).
N	vector or matrix of number of subjects.
Gamma	correlation matrix.
tol	numeric value specifying tolerance threshold for convergence.
par_size	numeric value specifying the number of workers for parallel computing (1 for sequential processing).

Details

The following are additional details describing the input arguments (for m SNPs/observations measured in d studies):

betas	m x d matrix.
sds	m x d matrix.
dfs	vector of length d or an m x d matrix.
alt_props	vector of length d.
mafs	vector of length m or an m x d matrix.
	If NULL, error variances will not be adjusted for MAF.
N	vector of length d or an m x d matrix.
	Must be specified if mafs!=NULL.
Gamma	d x d matrix.
	If NULL, will be estimated using observations where all |t| < 5.

Value

A list with the following elements:

post_prob	matrix of posterior probabilities (each column corresponds to an association pattern).
pis	vector of estimated proportion of observations belonging to each association pattern.
D_mat	matrix of densities under each association pattern.
Gamma	correlation matrix.
Tstat_mod	matrix of moderated t-statistics.
V_mat	matrix of scaling factors under the alternative distribution.
mdf_sd_mat	matrix of standard deviation adjustment according to moderated degrees of freedom:sqrt(df/(df-2)).
prior_df	vector of the prior degrees of freedom for each marginal distribution.
prior_var	vector of the prior variance estimators for each marginal distribution.
unscaled_var	vector of the unscaled variance priors on non-zero coefficients for each marginal distribution.

The main element of interest for inference is the posterior probabilities matrix, post_prob. The estimated proportion of observations belonging to each association pattern, pis, may also be of interest. The remaining elements are returned primarily for use by other functions.

kjgleason/primo documentation built on Sept. 7, 2021, 5:21 p.m.