Class MvnImputer

java.lang.Object

edu.cmu.tetrad.data.missing.MvnImputer

All Implemented Interfaces:

MultipleImputer

public final class MvnImputer
extends Object
implements MultipleImputer

A multiple imputer for continuous data under a saturated multivariate normal model. The mean and covariance are estimated once by EM (see EmCovarianceEstimator); each missing block is then drawn from its conditional normal distribution given the observed entries of its row: N(mu_M + S_MO S_OO^{-1} (x_O - mu_O), S_MM - S_MO S_OO^{-1} S_OM). Valid under MAR for approximately multivariate normal data.

Note: this is "improper" multiple imputation in Rubin's sense--the model parameters are fixed at their EM estimates rather than drawn from a posterior--so between-imputation variability is somewhat understated. For pooling graphs by edge frequency this is a reasonable first implementation; a proper (posterior-draw) variant is a flagged follow-up.

Version:

$Id: $Id

Author:

josephramsey

Constructor Summary

Constructors

Constructor	Description
MvnImputer()	Constructs an imputer with default EM settings.
MvnImputer(MissingDataSpec spec)	Constructs an imputer whose EM settings (ridge, tolerance, max iterations) are taken from the given spec.

Method Summary

Modifier and Type	Method	Description
List<DataSet>	impute(DataSet dataSet, int m, long seed)	Produces m completed copies of the given dataset.

Methods inherited from class Object

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

Constructor Details

MvnImputer

public MvnImputer()

Constructs an imputer with default EM settings.

MvnImputer

public MvnImputer(MissingDataSpec spec)

Constructs an imputer whose EM settings (ridge, tolerance, max iterations) are taken from the given spec.

Parameters:

spec - The spec.

Method Details

impute

public List<DataSet> impute(DataSet dataSet,
 int m,
 long seed)

Produces m completed copies of the given dataset.

Specified by:

impute in interface MultipleImputer

Parameters:

dataSet - The dataset; must contain at least one missing value.

m - The number of imputations; at least 1. With 1, the single completed dataset shows nothing of the uncertainty due to the missing values, which is what the differences among several show.

seed - A random seed for reproducibility, or -1 for a random seed.

Returns:

The m completed datasets.