e_step_func: Function for fitting the empirical Bayes portion of the...

In probe: Sparse High-Dimensional Linear Regression with PROBE

Function for fitting the empirical Bayes portion of the E-step

Description

A wrapper function estimating posterior expectations of the \gamma variables using an empirical Bayesian technqiue.

Usage

e_step_func(beta_t, beta_var, df, adj = 5, lambda = 0.1, monotone = TRUE)

Arguments

beta_t	Expectation of the posterior mean (assuming \gamma=1)
beta_var	Current posterior variance (assuming \gamma=1)
df	Degrees of freedom for the t-distribution (use to calculate p-values).
adj	Bandwidth multiplier to Silverman's ‘rule of thumb’ for calculating the marginal density of the test-statistics (default = 5).
lambda	Value of the \lambda parameter for estimating the proportion of null hypothesis using Storey et al. (2004) (default = 0.1).
monotone	Logical - Should the estimated marginal density of the test-statistics be monotone non-increasing from zero (default = TRUE).

Value

A list including

delta estimated posterior expectations of the \gamma.

pi0 estimated proportion of null hypothesis

References

Storey, J. D., Taylor, J. E., and Siegmund, D. (2004), “Strong control, conservative point estimation and simultaneous conservative consistency of false discovery rates: A unified approach,” J. R. Stat. Soc. Ser. B. Stat. Methodol., 66, 187–205. McLain, A. C., Zgodic, A., & Bondell, H. (2022). Sparse high-dimensional linear regression with a partitioned empirical Bayes ECM algorithm. arXiv preprint arXiv:2209.08139.

probe documentation built on Nov. 2, 2023, 5:49 p.m.