Class ImputationRegimeStudy

java.lang.Object
edu.cmu.tetrad.study.missingness.ImputationRegimeStudy
public final class ImputationRegimeStudy extends Object
Does MICE imputation with IMaGES-style pooling generalize beyond the linear Gaussian regime? Two phases, both BOSS at 20% MCAR missingness:
  • MIXED: Lee-Hastie simulation, half the variables discrete, scored by degenerate-Gaussian BIC. This is the cell where the missing-data literature's own simulations favored deletion over imputation in every scenario, so a MICE+IMaGES win here is the interesting outcome, not the expected one.
  • NONLINEAR: continuous data from an additive nonlinear SEM (each edge's function drawn from linear, quadratic, cubic, or tanh), scored by basis-function BIC at truncation 3. The hazard for imputation here is specific: MICE's donors are matched on a linear fitted mean, so the imputed cells' dependence is flattened toward linearity even while their margins stay honest, and the embedded score consumes exactly the dependence.

Arms per phase: oracle (complete data), listwise, test-wise deletion at fullN and at minPairwise, EM on the embedded matrix at meanPairwise (the covariance-estimating representative; for BF this breaks the deterministic relations among basis columns, a known caveat), and MICE m = 5 pooled IMaGES-style (one BOSS run over the average of the completed datasets' scores).

Run from IntelliJ. The DG phase is quick; the BF phase is the wall clock, since every IMaGES local score evaluates five basis-function scores. An arm that exceeds the timeout is abandoned for that rep and said so.

Author:
josephramsey

  • Constructor Summary

    Constructors
    Constructor
    Description
    ImputationRegimeStudy()
    Default constructor for the ImputationRegimeStudy class.

  • Method Summary

    Modifier and Type
    Method
    Description
    static void
    main(String[] args)
    The entry point: runs the mixed phase (degenerate-Gaussian score) and the nonlinear phase (basis-function score), printing a summary table per phase.

    Methods inherited from class Object

    equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

  • Constructor Details

    • ImputationRegimeStudy

      public ImputationRegimeStudy()
      Default constructor for the ImputationRegimeStudy class.

  • Method Details

    • main

      public static void main(String[] args) throws Exception
      The entry point: runs the mixed phase (degenerate-Gaussian score) and the nonlinear phase (basis-function score), printing a summary table per phase.
      Parameters:
      args - Command-line arguments; not used.
      Throws:
      Exception - If any error occurs during simulation or search.