noteargis: Learning a sparse DAG with Grouped Variables
In yongsu-lee/NOTEARGIS: Learn Directed Acyclic Graphs with Grouped Variables

Description Usage Arguments Value

View source: R/noteargis.R

Learning a sparse DAG with Grouped Variables

noteargis(
  data_input,
  data_type = c("c", "m"),
  n_resps_by_node = NULL,
  intcpt = NULL,
  lambdas = NULL,
  n_lams = NULL,
  admm_args = admm_arg_ctrl(),
  add_stop_rule = TRUE,
  fac_grp_lasso = FALSE,
  verbose = FALSE,
  fit_hist = FALSE
)

`data_input`	a data.frame with attributes `data_type` and either `n_levels` or `n_grp_elmts`. Check `NOTEARGIS::gen_data` for details.
`data_type`	a character vector, elements consist of either "c" or "m" indicating continuous variable and multinomial response, respectively.
`n_resps_by_node`	positive integers vector, a list of numbers of elements for each group or numbers of levels for each multinomial variable.
`intcpt`	intercept terms will be included if "always" and will be excluded if "none".
`lambdas`	a double vector, user specified tuning parameter sequence. Typical usage is to have the function compute its own list of tuning parameters based on `n_lams` (and internal function `gen_lambdas`). Do not recommend to supply this argument. Default is `NA`.
`n_lams`	a positive integer, the number of seuqences of tuning parameters. Default is `30`.
`admm_args`	a list of value, to customize the several arguments for ADMM algorithm. See `?admm_arg_ctrl` for details.
`add_stop_rule`	a logical scalar, if `add_stop_rule = TRUE`, the function will stop proceed fitting if the number of estimated edges are larger than `3*n_nodes`. Default is `TRUE`.
`fac_grp_lasso`	a logical scalar, if `fac_grp_lasso = TRUE`, square-root of number of parameters will be considered as the multiplied factor in the group lasso update. Default is `FALSE`.
`verbose`	a logical scalar, if `verbose = TRUE` you can see fitting status message in the console window. Default is `FALSE`.
`fit_hist`	a logical scalar, if `verbose = TRUE` function returns all the parameters at each step of the ADMM loop. Do not recommend. Default is `FALSE`.

An object with S3 class "noteargis"

`A_est_by_lam`	A list of estimated adjacency matrices over the pathwise solutions. `graph_est` is saved as an attribute for each pathwise solution.
`Beta_new_by_lam`	A list of Beta estimates over the pathwise solutions.
`W_new_by_lam`	A list of W estimates over the pathwise solutions.
`lambdas`	A list of tuning parameters used for fitting all the pathwise solutions
`history_W_by_lam`	(if `history = TRUE`) A list of W estimates for each step of ADMM over the pathwise solutions.
`history_Beta_by_lam`	(if `history = TRUE`) A list of Beta estimates for each step of ADMM over the pathwise solutions.