GAMupdate: Get an initial set of putative variables for the GAM...

In haimbar/SEMMS: Variable selection in Generalized Linear Models - an empirical Bayes approach

Description

Get an initial set of putative variables for the GAM algorithm

Usage

GAMupdate(
  initidx,
  initval,
  Yr,
  Xr,
  Z,
  distr = "N",
  randomize = TRUE,
  mincor = 0.7,
  maxsteps = 20L,
  minchange = 1,
  ptf = TRUE
)

Arguments

initidx	an initial set of variables to use in the fitting algorithm
initval	the values (-1 or 1) of the initial set of preditors
Yr	the response vector
Xr	the fixed-effect design matrix
Z	the matrix of all putative variables
distr	the GLM distribution to fit (N=Normal, B=binary, P=Poisson)
randomize	Boolean - whether to run the greedy or randomized version
mincor	a threshold of (absolute) correlation above which a pair is considered highly correlated
maxsteps	maximum number of GAM iterations
minchange	the minimum difference in log-likelihood between consecutive iterations below which we assume that the algorithm has converged
ptf	Boolean - whether to print debug messages to SEMMS.log

Value

a list containing the index of non-null variables (columns in Z), the mixture model parameters, the sign of the selected coefficients, a matrix with posterior probabilities, and an indicator array for locked out variables

haimbar/SEMMS documentation built on Dec. 20, 2021, 2:44 p.m.