Class DeterministicClusters.Fitter

java.lang.Object
edu.cmu.tetrad.data.audit.DeterministicClusters.Fitter
Enclosing class:
DeterministicClusters
public static final class DeterministicClusters.Fitter extends Object
A reusable equation fitter for one dataset: the correlation matrix and its ridge-regularized inverse are computed once, and each equation costs a few small solves. See the class Javadoc for the method.
Version:
$Id: $Id
Author:
josephramsey

  • Constructor Details

    • Fitter

      public Fitter(DataSet data, double threshold)
      Constructs the fitter, computing means, standard deviations, and the correlation matrix once.
      Parameters:
      data - The dataset.
      threshold - The exactness threshold; see DeterministicClusters.find(DataSet, double).

  • Method Details

    • leaveOneOut

      public DeterministicClusters.Constraint leaveOneOut(String varName)
      Fits the equation writing the named variable in terms of ALL other analyzed variables (screened to the ones that matter).
      Parameters:
      varName - The variable for the left-hand side.
      Returns:
      The fitted constraint; a constant variable comes back with empty support and residual fraction 0; null if the variable is not an analyzed or constant continuous variable of the dataset.

    • onCommonSupport

      public List<DeterministicClusters.Constraint> onCommonSupport(List<String> targetNames, Collection<String> supportNames)
      Fits, for each target, its equation in terms of the SAME support universe -- the audit's retained-form report, where every removed variable is written in terms of the retained variables. One inverse of the support submatrix serves all targets.
      Parameters:
      targetNames - The variables for the left-hand sides.
      supportNames - The common support universe (the targets themselves are excluded automatically).
      Returns:
      One constraint per target, in order; null entries for targets that are not analyzed continuous variables or when fewer than one support variable is analyzed.