Skip to contents

SuSiE with Regression Summary Statistics (RSS)

Source: R/susie.R
susie_rss.Rd

Performs SuSiE regression using z-scores and correlation matrix. Supports both standard RSS (lambda = 0) and RSS with regularized LD matrix (lambda > 0).

Usage

susie_rss(
  z = NULL,
  R,
  n = NULL,
  bhat = NULL,
  shat = NULL,
  var_y = NULL,
  L = min(10, ncol(R)),
  lambda = 0,
  maf = NULL,
  maf_thresh = 0,
  z_ld_weight = 0,
  prior_variance = 50,
  scaled_prior_variance = 0.2,
  residual_variance = NULL,
  prior_weights = NULL,
  null_weight = 0,
  standardize = TRUE,
  intercept_value = 0,
  estimate_residual_variance = FALSE,
  estimate_residual_method = c("MoM", "MLE"),
  estimate_prior_variance = TRUE,
  estimate_prior_method = c("optim", "EM", "simple"),
  unmappable_effects = c("none", "inf"),
  check_null_threshold = 0,
  prior_tol = 1e-09,
  residual_variance_lowerbound = 0,
  residual_variance_upperbound = Inf,
  model_init = NULL,
  coverage = 0.95,
  min_abs_corr = 0.5,
  max_iter = 100,
  tol = 0.001,
  convergence_method = c("elbo", "pip"),
  verbose = FALSE,
  track_fit = FALSE,
  check_input = FALSE,
  check_prior = TRUE,
  check_R = TRUE,
  check_z = FALSE,
  n_purity = 100,
  r_tol = 1e-08,
  refine = FALSE
)

Arguments

z

A p-vector of z-scores.

R

A p by p correlation matrix.

n

The sample size, not required but recommended.

bhat

Alternative summary data giving the estimated effects (a vector of length p). This, together with shat, may be provided instead of z.

shat

Alternative summary data giving the standard errors of the estimated effects (a vector of length p). This, together with bhat, may be provided instead of z.

lambda

Regularization parameter for LD matrix. When lambda > 0, you cannot use unmappable_effects methods.

z_ld_weight

This parameter is included for backwards compatibility with previous versions of the function, but it is no longer recommended to set this to a non-zero value. When z_ld_weight > 0, the matrix R is adjusted to be cov2cor((1-w)*R + w*tcrossprod(z)), where w = z_ld_weight.

estimate_residual_variance

The default is FALSE, the residual variance is fixed to 1 or variance of y. If the in-sample LD matrix is provided, we recommend setting estimate_residual_variance = TRUE.

check_R

If TRUE, check that R is positive semidefinite.

check_z

If TRUE, check that z lies in column space of R.