one_step_update: title one step update in flash iteration using ash

In kkdey/flashr: Factor Loading Adaptive SHrinkage in R

Description

description one step update in flash iteration using ash

Usage

one_step_update(Y, El, El2, Ef, Ef2, N, P, sigmae2_v, sigmae2_true, sigmae2,
  nonnegative = FALSE, partype = "constant", objtype = "margin_lik",
  fix_factor = FALSE, ash_para = list(), fl_list = list())

Arguments

Y	the data matrix
El	mean for the loadings
El2	second moment for the loadings
Ef	mean for the factors
Ef2	second moment for the factors
N	dimension of Y
P	dimension of Y
sigmae2_v	residual square
sigmae2_true	the (true) known variance structure Here, sigmae2 is the estimated variance structure in each step sigmae2_true is the truth we know, some times sigmae2 is noisy version of sigmae2_true
sigmae2	the estimation of the variance structure
nonnegative	if the facotor and loading are nonnegative or not. TRUE for nonnegative FALSE for no constraint
partype	parameter type for the variance, "constant" for constant variance, "var_col" for nonconstant variance for column, "known" for the kown variance, "Bayes_var" for Bayes version of the nonconstant variance for row and column "loganova" is anova estiamtion for the log residual square
objtype	objective function type, "margin_lik" for conditional likelihood, "lowerbound_lik" for full objective function
fix_factor	whether the factor is fixed or not TRUE for fix_factor FALSE for non-constraint

Value

list of factor, loading and variance of noise matrix

• El is a N vector for mean of loadings

• El2 is a N vector for second moment of loadings

• Ef is a N vector for mean of factors

• Ef2 is a N vector for second moment of factors

• sigmae2_vis a N by P matrix for residual square

• sigmae2_trueis a N by P matrix for estimated value for the variance structure

• obj_valthe value of objectice function

kkdey/flashr documentation built on May 20, 2019, 10:36 a.m.