AnovaVarianceEstimator.java

package umontreal.ssj.mcqmctools.anova;
 
import umontreal.ssj.rng.RandomStream;
import umontreal.ssj.hups.PointSetIterator;
import umontreal.ssj.mcqmctools.*;
import umontreal.ssj.stat.Tally;
 
import java.util.*;

public class AnovaVarianceEstimator implements MonteCarloModel<double[]> {
 
   // partial variance estimator
   protected PartialVarianceEstimator varEstimator;
 
   // partial variance integrator
   protected RandomIntegrator integrator;
 
   // variance storage
   double[] vars = null;
 
   public AnovaVarianceEstimator() {
      this.integrator = null;
      this.varEstimator = new PartialVarianceEstimator();
   }
 
   public MonteCarloModelDouble getModel() {
      return varEstimator.getModel();
   }
 
   public double getApproximateMean() {
      return varEstimator.getApproximateMean();
   }
 
   public void setModel(MonteCarloModelDoubleRQMC model, double approxMean) {
      varEstimator.setModel(model, approxMean);
   }
 
   public Integrator getIntegrator() {
      return integrator;
   }

   public void setIntegrator(RandomIntegrator integrator) {
      this.integrator = integrator;
   }

   public List<CoordinateSet> getCoordinateSets() {
      return varEstimator.getCoordinateSets();
   }

   public void setCoordinates(List<CoordinateSet> coordSets) {
      varEstimator.setCoordinateSets(coordSets);
   }

   public void setCoordinates(CoordinateSet coords) {
      varEstimator.setCoordinateSets(coords.subsetsNotEmpty());
   }

   public void setCoordinates(CoordinateSet coords, int maxOrder) {
      varEstimator.setCoordinateSets(coords.subsetsNotEmpty(maxOrder));
   }

   public void simulate(RandomStream stream) {
 
      List<CoordinateSet> coordSets = varEstimator.getCoordinateSets();
      int nSets = coordSets.size();
 
      if (vars == null || vars.length != nSets + 2)
         vars = new double[nSets + 2];
 
      if (integrator == null)
         throw new IllegalStateException("integrator has not been initialized");
 
      integrator.setStream(stream);
      integrator.integrate(varEstimator, vars);
 
      vars[nSets + 1] -= vars[nSets] * vars[nSets];
 
      for (int i = 0; i < nSets; i++) {
 
         CoordinateSet cs = coordSets.get(i);
 
         // the components are ordered by coordinate mask
         // so there is no need to iterate over all of them
         for (int i2 = 0; i2 < i; i2++) {
 
            CoordinateSet cs2 = coordSets.get(i2);
 
            if (cs2.isSubset(cs))
               vars[i] -= vars[i2];
         }
      }
   }
 
   public double[] getPerformance() {
      return vars;
   }

   public int getDimension() {
      return varEstimator.getDimension();
   }

   @Override
   public String toString() {
      String s = String.format("ANOVA Variance Estimator" + " [model=%s]", varEstimator.getModel());
      return s;
   }

   public String getTag() {
      return getModel().getTag();
   }
   
}