The StARS tuning parameter selection method from package 'huge'. Adapted to JGL models.

The StARS tuning parameter selection method from package 'huge'. Adapted to JGL models.

huge_stars_joint(
  inputlist,
  lambda = NULL,
  nlambda = 10,
  lambda.min.ratio = NULL,
  stars.subsample.ratio.vec = NULL,
  rep.num = 20,
  stars.thresh = 0.1,
  lam2init = NULL,
  lam1init = NULL,
  verbose = T,
  ...
)

Arguments

inputlist	a list of matrices with Gaussian distributed counts
lambda	a vector of candidate tuning parameter values. If NULL, the values will be generated according to the data.
nlambda	interger, number of candidate lambda values
lambda.min.ratio	numeric, the ratio of lambda.min / lambda.max, if NULL, it willbe set to 0.1
stars.subsample.ratio.vec	vector, the subsample ratios. If NULL, it will be decided by data
rep.num	integer, the number of subsampling
stars.thresh	the threshold value in StARS
lam2init	the initial value of lambda2 in JGL. This is set when try to optimize over lambda1
lam1init	the initial value of lambda1 in JGL. This is set when try to optimize over lambda2
verbose	logical, whether show processing messages.

Value

lambda: a vector of candidate lambda used. opt.lambda: the selected lambda value