Merge branch 'release-1.0.23'

Author: tesar-stepan

lmtest.weights1.txt

@@ -5,7 +5,7 @@
 	<name>baseCode</name>
 	<groupId>baseCode</groupId>
 	<artifactId>baseCode</artifactId>
-	<version>1.0.22</version>
+	<version>1.0.23</version>
 	<inceptionYear>2003</inceptionYear>
 	<description>
 	<![CDATA[Data structures, math and statistics tools, and utilities that are often needed across projects.]]>

pom.xml

@@ -1,8 +1,8 @@
 /*
  * The baseCode project
- * 
+ *
  * Copyright (c) 2006 University of British Columbia
- * 
+ *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
  * You may obtain a copy of the License at
@@ -21,15 +21,14 @@
 import java.util.List;
 import java.util.Vector;
 
+import cern.colt.list.IntArrayList;
 import ubic.basecode.dataStructure.matrix.Matrix2D;
 import ubic.basecode.dataStructure.matrix.MatrixUtil;
-import cern.colt.list.IntArrayList;
 
 /**
  * Remove rows from a matrix that are missing too many points.
- * 
+ *
  * @author Paul Pavlidis
- * @version $Id$
  */
 public class RowMissingFilter<M extends Matrix2D<R, C, V>, R, C, V> extends AbstractFilter<M, R, C, V> {
 
@@ -42,8 +41,8 @@ public class RowMissingFilter<M extends Matrix2D<R, C, V>, R, C, V> extends Abst
 
     @Override
     public M filter( M data ) {
-        List<V[]> MTemp = new Vector<V[]>();
-        List<R> rowNames = new Vector<R>();
+        List<V[]> MTemp = new Vector<>();
+        List<R> rowNames = new Vector<>();
         int numRows = data.rows();
         int numCols = data.columns();
         IntArrayList present = new IntArrayList( numRows );
@@ -117,8 +116,9 @@ public M filter( M data ) {
         returnval.setColumnNames( data.getColNames() );
         returnval.setRowNames( rowNames );
 
-        log.info( "There are " + kept + " rows after removing rows which have fewer than " + minPresentCount
-                + " values (or fewer than " + ABSOLUTEMINPRESENT + ")" );
+        if ( kept < numRows )
+            log.info( "There are " + kept + " rows after removing rows which have fewer than " + Math.max( ABSOLUTEMINPRESENT, minPresentCount )
+                    + " values" );
 
         return returnval;
 
@@ -127,7 +127,7 @@ public M filter( M data ) {
     /**
      * Set the maximum fraction of rows which will be removed from the data set. The default value is 0.3 Set it to 1.0
      * to remove this restriction.
-     * 
+     *
      * @param f double
      */
     public void setMaxFractionRemoved( double f ) {
@@ -140,7 +140,7 @@ public void setMaxFractionRemoved( double f ) {
      * Set the minimum number of values that must be present in each row. The default value is 5. This is always
      * overridden by a hard-coded value (currently 2) that must be present for a row to be kept; but this value is in
      * turn overridden by the maxfractionRemoved.
-     * 
+     *
      * @param m int
      */
     public void setMinPresentCount( int m ) {

src/ubic/basecode/datafilter/RowMissingFilter.java

@@ -127,7 +127,7 @@ public void setTopLeft( String topLeft ) {
     /**
      * @param <V>
      * @param matrix
-     * @param printNames Should the row and column names be included
+     * @param printNames Should the row and column names be included; FIXME this fails if names aren't provided
      * @throws IOException
      */
     public <V> void writeMatrix( Matrix2D<R, C, V> matrix, boolean printNames ) throws IOException {

src/ubic/basecode/io/writer/MatrixWriter.java

@@ -154,7 +154,7 @@ public static int[] chooserandomWrep( int max, int n ) {
     }
 
     /**
-     * Initialized the random number generator witha given seed.
+     * Initialized the random number generator with a given seed.
      * 
      * @param seed
      */

src/ubic/basecode/math/RandomChooser.java

@@ -92,6 +92,8 @@ public class QRDecomposition {
      */
     private DoubleMatrix2D Qcached = null;
 
+    private DoubleMatrix2D effects;
+
     /**
      * @param A the matrix to decompose, pivoting will be used.
      */
@@ -227,39 +229,41 @@ public DoubleMatrix2D chol2inv() {
     }
 
     /**
-     * Compute effects matrix Q'y (as in Rb = Q'y). Returns only
-     * the aspect associated with the parameters - the first <tt>rank</tt> elements.
+     * Compute effects matrix Q'y (as in Rb = Q'y).
+     * 
+     * <p>
+     * "Tthe effects are the uncorrelated single-degree-of-freedom values obtained by projecting the data onto the
+     * successive orthogonal subspaces generated by the QR decomposition during the fitting process. The first r (the
+     * rank of the model) are associated with coefficients and the remainder span the space of residuals (but are not
+     * associated with particular residuals)."
      * 
      * @param y vector Missing values are ignored, otherwise assumed to be of the right size
      * @return vector of effects - these are the projections of y into Q column space
      */
     public DoubleMatrix1D effects( DoubleMatrix1D y ) {
 
-        Algebra solver = new Algebra();
-        // Inefficient due to getting Q explicitly
-        return solver.mult( solver.transpose( this.getQ() ), MatrixUtil.removeMissing( y ) ).viewPart( 0, this.rank );
-
-        // Could do it this way, but QR.toArray() makes a copy...
-        //        double[] qty = new double[y.size()];
-        //        double[] junk = new double[y.size()];
-        //        ubic.basecode.math.linalg.Dqrsl.dqrsl_j( QR.toArray(), QR.rows(), QR.columns(), qraux.toArray(), y.toArray(),
-        //                junk, qty,
-        //                junk, junk, junk, 1000 ); 
-        //        return new DenseDoubleMatrix1D( qty );
+        double[] qty = new double[y.size()];
+        double[] junk = new double[y.size()];
+        ubic.basecode.math.linalg.Dqrsl.dqrsl_j( QR.toArray(), QR.rows(), QR.columns(), qraux.toArray(), MatrixUtil.removeMissing( y ).toArray(),
+                junk, qty,
+                junk, junk, junk, 1000 );
+        return new DenseDoubleMatrix1D( qty );
     }
 
     /**
-     * Compute effects matrix Q'y (as in Rb = Q'y) Returns only the aspect associated with the parameters - the first
-     * <tt>rank</tt> elements.
+     * Compute effects matrix Q'y (as in Rb = Q'y)
      * 
      * @param y matrix of data, assumed to be of right size, missing values are not supported
      * @return matrix of effects - these are the projections of y's columns into Q column subspace associated with the
      *         parameters,
      *         so values are "effects" each basis vector on the data
      */
     public DoubleMatrix2D effects( DoubleMatrix2D y ) {
-        Algebra solver = new Algebra();
-        return solver.mult( solver.transpose( this.getQ() ), y );
+        double[][] efa = new double[y.columns()][y.rows()];
+        for ( int i = 0; i < y.columns(); i++ ) {
+            efa[i] = effects( y.viewColumn( i ) ).toArray();
+        }
+        return new DenseDoubleMatrix2D( efa ).viewDice();
     }
 
     /**
@@ -387,11 +391,7 @@ public DoubleMatrix2D solve( DoubleMatrix2D y ) {
             throw new IllegalArgumentException( "Matrix is rank deficient; try using pivoting" );
         }
 
-        /*
-         * Direct computation of Q'y can be avoided by manipulation of the compact QR instead (see dqrsl).
-         * 
-         */
-        DoubleMatrix2D qTy = effects( y ); // will be modified by this call.
+        DoubleMatrix2D qTy = effects( y ); // FIXME we use this again later, but we recompute it. Try to cache it.
 
         // Solve R*X = Y => X = RinvY; backsubstitution
         for ( int k1 = rank - 1; k1 >= 0; k1-- ) {
@@ -421,7 +421,7 @@ public DoubleMatrix2D solve( DoubleMatrix2D y ) {
          */
         DoubleMatrix2D coeff = qTy.like( p, y.columns() );
         coeff.assign( Double.NaN );
-        for ( int i = 0; i < qTy.rows(); i++ ) {
+        for ( int i = 0; i < this.rank; i++ ) {
             int piv = jpvt[i]; // where the value should go.
             for ( int j = 0; j < qTy.columns(); j++ ) {
                 coeff.setQuick( piv, j, qTy.getQuick( i, j ) );