MDFS {MDFS}

Run end-to-end MDFS

Description

Run end-to-end MDFS

Usage

MDFS(
  data,
  decision,
  n.contrast = max(ncol(data), 30),
  dimensions = 1,
  divisions = 1,
  discretizations = 1,
  range = NULL,
  pc.xi = 0.25,
  p.adjust.method = "holm",
  level = 0.05,
  seed = NULL,
  use.CUDA = FALSE
)

Arguments

data

input data where columns are variables and rows are observations (all numeric)

decision

decision variable as a boolean vector of length equal to number of observations

n.contrast

number of constrast variables (defaults to max of 1/10 of variables number and 30)

dimensions

number of dimensions (a positive integer; on CUDA limited to 2–5 range)

divisions

number of divisions (from 1 to 15)

discretizations

number of discretizations

range

discretization range (from 0.0 to 1.0; NULL selects probable optimal number)

pc.xi

parameter xi used to compute pseudocounts (the default is recommended not to be changed)

p.adjust.method

method as accepted by p.adjust ("BY" is recommended for FDR, see Details)

level

statistical significance level

seed

seed for PRNG used during discretizations (NULL for random)

use.CUDA

whether to use CUDA acceleration (must be compiled with CUDA; NOTE: the CUDA version might provide a slightly lower sensitivity due to a lack of native support for contrast_data)

Details

In case of FDR control it is recommended to use Benjamini-Hochberg-Yekutieli p-value adjustment method ("BY" in p.adjust) due to unknown dependencies between tests.

Value

A list with the following fields:

• contrast.indices – indices of variables chosen to build contrast variables

• contrast.variables – built contrast variables

• MIG.Result – result of ComputeMaxInfoGains

• MDFS – result of ComputePValue (the MDFS object)

• statistic – vector of statistic's values (IGs) for corresponding variables

• p.value – vector of p-values for corresponding variables

• adjusted.p.value – vector of adjusted p-values for corresponding variables

• relevant.variables – vector of relevant variables indices

Examples

MDFS(madelon$data, madelon$decision, dimensions = 2, divisions = 1,
     range = 0, seed = 0)