Examples with TetradVector edu.cmu.tetrad.util.TetradVector used on opensource projects

Example 21 with TetradVector

use of edu.cmu.tetrad.util.TetradVector in project tetrad by cmu-phil.

the class SemBicScoreDeterministic method getMaximalLinearlyDependentSet.

private int[] getMaximalLinearlyDependentSet(int i, int[] parents, ICovarianceMatrix cov) {
    double small = getDeterminismThreshold();
    List<Node> _parents = new ArrayList<>();
    for (int p : parents) _parents.add(variables.get(p));
    DepthChoiceGenerator gen = new DepthChoiceGenerator(_parents.size(), _parents.size());
    int[] choice;
    while ((choice = gen.next()) != null) {
        int[] sel0 = new int[choice.length];
        List<Integer> all = new ArrayList<>();
        for (int w = 0; w < parents.length; w++) all.add(parents[w]);
        for (int w = 0; w < sel0.length; w++) all.remove(sel0[w]);
        int[] sel = new int[all.size()];
        for (int w = 0; w < all.size(); w++) sel[w] = all.get(w);
        List<Node> _sel = new ArrayList<>();
        for (int m = 0; m < choice.length; m++) {
            sel[m] = parents[m];
            _sel.add(variables.get(sel[m]));
        }
        TetradMatrix m = cov.getSelection(sel, sel);
        double s2 = getCovariances().getValue(i, i);
        TetradMatrix covxx = getSelection(getCovariances(), parents, parents);
        TetradVector covxy = getSelection(getCovariances(), parents, new int[] { i }).getColumn(0);
        s2 -= covxx.inverse().times(covxy).dotProduct(covxy);
        if (s2 <= small) {
            out.println("### Linear dependence among variables: " + _sel);
            out.println("### Removing " + _sel.get(0));
            return sel;
        }
        try {
            m.inverse();
        } catch (Exception e2) {
            // forbidden.add(sel[0]);
            out.println("### Linear dependence among variables: " + _sel);
            out.println("### Removing " + _sel.get(0));
            return sel;
        }
    }
    return new int[0];
}

Example 22 with TetradVector

use of edu.cmu.tetrad.util.TetradVector in project tetrad by cmu-phil.

the class SemBicScoreImages2 method localScore.

/**
 * Calculates the sample likelihood and BIC score for i given its parents in a simple SEM model
 */
public double localScore(int i, int... parents) {
    for (int p : parents) if (forbidden.contains(p))
        return Double.NaN;
    double lik = 0.0;
    for (int k = 0; k < covariances.size(); k++) {
        double residualVariance = getCovariances(k).getValue(i, i);
        TetradMatrix covxx = getSelection1(getCovariances(k), parents);
        try {
            TetradMatrix covxxInv = covxx.inverse();
            TetradVector covxy = getSelection2(getCovariances(k), parents, i);
            TetradVector b = covxxInv.times(covxy);
            residualVariance -= covxy.dotProduct(b);
            if (residualVariance <= 0) {
                if (isVerbose()) {
                    out.println("Nonpositive residual varianceY: resVar / varianceY = " + (residualVariance / getCovariances(k).getValue(i, i)));
                }
                return Double.NaN;
            }
            int cols = getCovariances(0).getDimension();
            double q = 2 / (double) cols;
            lik += -sampleSize * Math.log(residualVariance);
        } catch (Exception e) {
            boolean removedOne = true;
            while (removedOne) {
                List<Integer> _parents = new ArrayList<>();
                for (int y = 0; y < parents.length; y++) _parents.add(parents[y]);
                _parents.removeAll(forbidden);
                parents = new int[_parents.size()];
                for (int y = 0; y < _parents.size(); y++) parents[y] = _parents.get(y);
                removedOne = printMinimalLinearlyDependentSet(parents, getCovariances(k));
            }
            return Double.NaN;
        }
    }
    int p = parents.length;
    double c = getPenaltyDiscount();
    return 2 * lik - c * (p + 1) * Math.log(covariances.size() * sampleSize);
}

Example 23 with TetradVector

use of edu.cmu.tetrad.util.TetradVector in project tetrad by cmu-phil.

the class RegressionCovariance method regress.

/**
 * Regresses the given target on the given regressors, yielding a regression
 * plane, in which coefficients are given for each regressor plus the
 * constant (if means have been specified, that is, for the last), and se,
 * t, and p values are given for each regressor.
 *
 * @param target     The variable being regressed.
 * @param regressors The list of regressors.
 * @return the regression plane.
 */
public RegressionResult regress(Node target, List<Node> regressors) {
    TetradMatrix allCorrelations = correlations.getMatrix();
    List<Node> variables = correlations.getVariables();
    int yIndex = variables.indexOf(target);
    int[] xIndices = new int[regressors.size()];
    for (int i = 0; i < regressors.size(); i++) {
        xIndices[i] = variables.indexOf(regressors.get(i));
        if (xIndices[i] == -1) {
            throw new NullPointerException("Can't find variable " + regressors.get(i) + " in this list: " + variables);
        }
    }
    TetradMatrix rX = allCorrelations.getSelection(xIndices, xIndices);
    TetradMatrix rY = allCorrelations.getSelection(xIndices, new int[] { yIndex });
    TetradMatrix bStar = rX.inverse().times(rY);
    TetradVector b = new TetradVector(bStar.rows() + 1);
    for (int k = 1; k < b.size(); k++) {
        double sdY = sd.get(yIndex);
        double sdK = sd.get(xIndices[k - 1]);
        b.set(k, bStar.get(k - 1, 0) * (sdY / sdK));
    }
    b.set(0, Double.NaN);
    if (means != null) {
        double b0 = means.get(yIndex);
        for (int i = 0; i < xIndices.length; i++) {
            b0 -= b.get(i + 1) * means.get(xIndices[i]);
        }
        b.set(0, b0);
    }
    int[] allIndices = new int[1 + regressors.size()];
    allIndices[0] = yIndex;
    for (int i = 1; i < allIndices.length; i++) {
        allIndices[i] = variables.indexOf(regressors.get(i - 1));
    }
    TetradMatrix r = allCorrelations.getSelection(allIndices, allIndices);
    TetradMatrix rInv = r.inverse();
    int n = correlations.getSampleSize();
    int k = regressors.size() + 1;
    double vY = rInv.get(0, 0);
    double r2 = 1.0 - (1.0 / vY);
    // Book says n - 1.
    double tss = n * sd.get(yIndex) * sd.get(yIndex);
    double rss = tss * (1.0 - r2);
    double seY = Math.sqrt(rss / (double) (n - k));
    TetradVector sqErr = new TetradVector(allIndices.length);
    TetradVector t = new TetradVector(allIndices.length);
    TetradVector p = new TetradVector(allIndices.length);
    sqErr.set(0, Double.NaN);
    t.set(0, Double.NaN);
    p.set(0, Double.NaN);
    TetradMatrix rxInv = rX.inverse();
    for (int i = 0; i < regressors.size(); i++) {
        double _r2 = 1.0 - (1.0 / rxInv.get(i, i));
        double _tss = n * sd.get(xIndices[i]) * sd.get(xIndices[i]);
        double _se = seY / Math.sqrt(_tss * (1.0 - _r2));
        double _t = b.get(i + 1) / _se;
        double _p = 2 * (1.0 - ProbUtils.tCdf(Math.abs(_t), n - k));
        sqErr.set(i + 1, _se);
        t.set(i + 1, _t);
        p.set(i + 1, _p);
    }
    // Graph
    this.graph = createGraph(target, allIndices, regressors, p);
    String[] vNames = createVarNamesArray(regressors);
    double[] bArray = b.toArray();
    double[] tArray = t.toArray();
    double[] pArray = p.toArray();
    double[] seArray = sqErr.toArray();
    return new RegressionResult(false, vNames, n, bArray, tArray, pArray, seArray, r2, rss, alpha, null, null);
}

Example 24 with TetradVector

use of edu.cmu.tetrad.util.TetradVector in project tetrad by cmu-phil.

the class RegressionDatasetGeneralized method regress.

/**
 * Regresses the target on the given regressors.
 *
 * @param target     The target variable.
 * @param regressors The regressor variables.
 * @return The regression plane, specifying for each regressors its
 * coefficeint, se, t, and p values, and specifying the same for the
 * constant.
 */
public RegressionResult regress(Node target, List<Node> regressors) {
    int n = data.rows();
    int k = regressors.size() + 1;
    int _target = variables.indexOf(target);
    int[] _regressors = new int[regressors.size()];
    for (int i = 0; i < regressors.size(); i++) {
        _regressors[i] = variables.indexOf(regressors.get(i));
    }
    int[] rows = new int[data.rows()];
    for (int i = 0; i < rows.length; i++) rows[i] = i;
    // TetradMatrix y = data.viewSelection(rows, new int[]{_target}).copy();
    TetradMatrix xSub = data.getSelection(rows, _regressors);
    // TetradMatrix y = data.subsetColumns(Arrays.asList(target)).getDoubleData();
    // RectangularDataSet rectangularDataSet = data.subsetColumns(regressors);
    // TetradMatrix xSub = rectangularDataSet.getDoubleData();
    TetradMatrix X = new TetradMatrix(xSub.rows(), xSub.columns() + 1);
    for (int i = 0; i < X.rows(); i++) {
        for (int j = 0; j < X.columns(); j++) {
            if (j == 0) {
                X.set(i, j, 1);
            } else {
                X.set(i, j, xSub.get(i, j - 1));
            }
        }
    }
    // for (int i = 0; i < zList.size(); i++) {
    // zCols[i] = getVariable().indexOf(zList.get(i));
    // }
    // int[] zRows = new int[data.rows()];
    // for (int i = 0; i < data.rows(); i++) {
    // zRows[i] = i;
    // }
    TetradVector y = data.getColumn(_target);
    TetradMatrix Xt = X.transpose();
    TetradMatrix XtX = Xt.times(X);
    TetradMatrix G = XtX.inverse();
    TetradMatrix GXt = G.times(Xt);
    TetradVector b = GXt.times(y);
    TetradVector yPred = X.times(b);
    // TetradVector xRes = yPred.copy().assign(y, Functions.minus);
    TetradVector xRes = yPred.minus(y);
    double rss = rss(X, y, b);
    double se = Math.sqrt(rss / (n - k));
    double tss = tss(y);
    double r2 = 1.0 - (rss / tss);
    // TetradVector sqErr = TetradVector.instance(y.columns());
    // TetradVector t = TetradVector.instance(y.columns());
    // TetradVector p = TetradVector.instance(y.columns());
    // 
    // for (int i = 0; i < 1; i++) {
    // double _s = se * se * xTxInv.get(i, i);
    // double _se = Math.sqrt(_s);
    // double _t = b.get(i) / _se;
    // double _p = 2 * (1.0 - ProbUtils.tCdf(Math.abs(_t), n - k));
    // 
    // sqErr.set(i, _se);
    // t.set(i, _t);
    // p.set(i, _p);
    // }
    // 
    // this.graph = createOutputGraph(target.getNode(), y, regressors, p);
    // 
    String[] vNames = new String[regressors.size()];
    for (int i = 0; i < regressors.size(); i++) {
        vNames[i] = regressors.get(i).getName();
    }
    return new RegressionResult(false, vNames, n, b.toArray(), new double[0], new double[0], new double[0], r2, rss, alpha, yPred, xRes);
}

Example 25 with TetradVector

use of edu.cmu.tetrad.util.TetradVector in project tetrad by cmu-phil.

the class RegressionUtils method residuals.

public static DataSet residuals(DataSet dataSet, Graph graph) {
    Regression regression = new RegressionDataset(dataSet);
    TetradMatrix residuals = new TetradMatrix(dataSet.getNumRows(), dataSet.getNumColumns());
    for (int i = 0; i < dataSet.getNumColumns(); i++) {
        Node target = dataSet.getVariable(i);
        Node _target = graph.getNode(target.getName());
        if (_target == null) {
            throw new IllegalArgumentException("Data variable not in graph: " + target);
        }
        Set<Node> _regressors = new HashSet<>(graph.getParents(_target));
        System.out.println("For " + target + " regressors are " + _regressors);
        List<Node> regressors = new LinkedList<>();
        for (Node node : _regressors) {
            regressors.add(dataSet.getVariable(node.getName()));
        }
        RegressionResult result = regression.regress(target, regressors);
        TetradVector residualsColumn = result.getResiduals();
        // residuals.viewColumn(i).assign(residualsColumn);
        residuals.assignColumn(i, residualsColumn);
    }
    return ColtDataSet.makeContinuousData(dataSet.getVariables(), residuals);
}