Samples from the posterior distribution
posterior_samples.RdSamples from the posterior distribution
Usage
posterior_samples(
  beta,
  se,
  eaf,
  R,
  maxsize,
  tau0,
  r0,
  niter,
  burnin,
  p,
  seed = 456,
  excl.burnin = TRUE,
  a0 = 1,
  b0 = NULL,
  inds0 = NULL,
  standardize = TRUE,
  verbose = TRUE,
  clump = TRUE,
  clump_r2 = 0.99^2,
  check_ld = FALSE,
  ala = NULL
)Arguments
- beta
- Vector of effect sizes. 
- se
- Vector of standard errors. 
- eaf
- Vector of effect allele frequencies. 
- R
- LD matrix. 
- maxsize
- The maximum number of causal variants. 
- tau0
- Parameter tau. 
- r0
- Parameter $r$. 
- niter
- Number of iterations. 
- burnin
- Number of burn-in samples. 
- p
- Number of variants. 
- seed
- Random seed. 
- excl.burnin
- Should the burn-in be excluded? 
- a0
- Hyperparameter a for the model size prior. 
- b0
- Hyperparameter b for the model size prior. 
- inds0
- Initial model indices (not used). 
- standardize
- Should the effect sizes be standardised? Defaults to TRUE. 
- verbose
- Verbose output. 
- clump
- Whether to clump extremely highly correlated variants. 
- clump_r2
- Clumping threshold for extremely highly correlated variants. 
- check_ld
- Should the Z-scores be checked for LD discrepancy? 
- ala
- Whether Approximate Laplace should be used?