aslminternal {AdaptiveSparsity}  R Documentation 
These are the fitting and initialization functions used by aslm. These should generally not be used.
figEM(x, y, init = NULL, stopDiff = 1e08, epsilon = 1e06, a = 1) fit.ols.lm(x, y) init.ones(x, y) init.rnorm(x, y) init.runif(x, y)
x 
design matrix of dimension 
y 
vector of observations of length n, or a matrix with n rows. 
init 
optional initialization, a list with components containing an initial estimate for 
stopDiff 
convergence criteria. Algorithm stops once difference in beta and sigma from one iteration to the next is less than stopDiff. 
epsilon 
amount to add to beta for numerical stability, 
a 
scaling of sigmaSqr to provide numerical stability for solving steps. 
figEM computes the Figueiredo EM algorithm for adaptive sparsity using Jeffreys prior.
fit.ols.lm computes an initial beta and sigma based on finding the lm.fit of the full design matrix.
init.ones computes an initial beta that is all ones and computes the associated sigmas.
init.rnorm computes an initial beta that is normally distributed with a mean of 0 and a standard deviation of 50
init.runif computes an initial beta that is uniformly distributed from 0 to 1
Currently, figEM uses fit.ols.lm to initialize beta and sigma if no init list is provided.
figEM returns a list with the following components:
coefficients 

vcov 
variancecovariance matrix. 
sigma 
norm of the model error. 
df 
degrees of freedom of residuals. 
fit.ols.lm and init.ones are used to initialize beta and sigma if init is not provided to figEM. Each of these functions returns a list with the following components:
beta 
initial 
sigma 
initial norm of the model error based on this initial beta. 
Figueiredo, M.A.T.; , “Adaptive sparseness for supervised learning”, Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.25, no.9, pp. 1150 1159, Sept. 2003
aslm
, which should be used directly instead of these methods