weka.experiment.PairedTTester Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of weka-dev Show documentation
Show all versions of weka-dev Show documentation
The Waikato Environment for Knowledge Analysis (WEKA), a machine
learning workbench. This version represents the developer version, the
"bleeding edge" of development, you could say. New functionality gets added
to this version.
/*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see
* -D <index,index2-index4,...>
* Specify list of columns that specify a unique
* dataset.
* First and last are valid indexes. (default none)
*
*
*
* -R <index>
* Set the index of the column containing the run number
*
*
*
* -F <index>
* Set the index of the column containing the fold number
*
*
*
* -G <index1,index2-index4,...>
* Specify list of columns that specify a unique
* 'result generator' (eg: classifier name and options).
* First and last are valid indexes. (default none)
*
*
*
* -S <significance level>
* Set the significance level for comparisons (default 0.05)
*
*
*
* -V
* Show standard deviations
*
*
*
* -L
* Produce table comparisons in Latex table format
*
*
*
* -csv
* Produce table comparisons in CSV table format
*
*
*
* -html
* Produce table comparisons in HTML table format
*
*
*
* -significance
* Produce table comparisons with only the significance values
*
*
*
* -gnuplot
* Produce table comparisons output suitable for GNUPlot
*
*
*
*
* @author Len Trigg ([email protected])
* @version $Revision: 11542 $
*/
public class PairedTTester implements OptionHandler, Tester, RevisionHandler {
/** for serialization */
static final long serialVersionUID = 8370014624008728610L;
/** The set of instances we will analyse */
protected Instances m_Instances;
/** The index of the column containing the run number */
protected int m_RunColumn = 0;
/** The option setting for the run number column (-1 means last) */
protected int m_RunColumnSet = -1;
/** The option setting for the fold number column (-1 means none) */
protected int m_FoldColumn = -1;
/** The column to sort on (-1 means default sorting) */
protected int m_SortColumn = -1;
/** The sorting of the datasets (according to the sort column) */
protected int[] m_SortOrder = null;
/** The sorting of the columns (test base is always first) */
protected int[] m_ColOrder = null;
/** The significance level for comparisons */
protected double m_SignificanceLevel = 0.05;
/**
* The range of columns that specify a unique "dataset" (eg: scheme plus
* configuration)
*/
protected Range m_DatasetKeyColumnsRange = new Range();
/** An array containing the indexes of just the selected columns */
protected int[] m_DatasetKeyColumns;
/** The list of dataset specifiers */
protected DatasetSpecifiers m_DatasetSpecifiers = new DatasetSpecifiers();
/**
* The range of columns that specify a unique result set (eg: scheme plus
* configuration)
*/
protected Range m_ResultsetKeyColumnsRange = new Range();
/** An array containing the indexes of just the selected columns */
protected int[] m_ResultsetKeyColumns;
/** An array containing the indexes of the datasets to display */
protected int[] m_DisplayedResultsets = null;
/** Stores a vector for each resultset holding all instances in each set */
protected ArrayList m_Resultsets = new ArrayList();
/** Indicates whether the instances have been partitioned */
protected boolean m_ResultsetsValid;
/** Indicates whether standard deviations should be displayed */
protected boolean m_ShowStdDevs = false;
/** the instance of the class to produce the output. */
protected ResultMatrix m_ResultMatrix = new ResultMatrixPlainText();
/** A list of unique "dataset" specifiers that have been observed */
protected class DatasetSpecifiers implements RevisionHandler, Serializable {
/** for serialization. */
private static final long serialVersionUID = -9020938059902723401L;
/** the specifiers that have been observed */
ArrayList m_Specifiers = new ArrayList();
/**
* Removes all specifiers.
*/
protected void removeAllSpecifiers() {
m_Specifiers.clear();
}
/**
* Add an instance to the list of specifiers (if necessary)
*
* @param inst the instance to add
*/
protected void add(Instance inst) {
for (int i = 0; i < m_Specifiers.size(); i++) {
Instance specifier = m_Specifiers.get(i);
boolean found = true;
for (int m_DatasetKeyColumn : m_DatasetKeyColumns) {
if (inst.value(m_DatasetKeyColumn) != specifier
.value(m_DatasetKeyColumn)) {
found = false;
}
}
if (found) {
return;
}
}
m_Specifiers.add(inst);
}
/**
* Get the template at the given position.
*
* @param i the index
* @return the template
*/
protected Instance specifier(int i) {
return m_Specifiers.get(i);
}
/**
* Gets the number of specifiers.
*
* @return the current number of specifiers
*/
protected int numSpecifiers() {
return m_Specifiers.size();
}
/**
* Returns the revision string.
*
* @return the revision
*/
@Override
public String getRevision() {
return RevisionUtils.extract("$Revision: 11542 $");
}
}
/** Utility class to store the instances pertaining to a dataset */
protected class Dataset implements RevisionHandler, Serializable {
/** for serialization. */
private static final long serialVersionUID = -2801397601839433282L;
/** the template */
Instance m_Template;
/** the dataset */
ArrayList m_Dataset;
/**
* Constructor
*
* @param template the template
*/
public Dataset(Instance template) {
m_Template = template;
m_Dataset = new ArrayList();
add(template);
}
/**
* Returns true if the two instances match on those attributes that have
* been designated key columns (eg: scheme name and scheme options)
*
* @param first the first instance
* @return true if first and second match on the currently set key columns
*/
protected boolean matchesTemplate(Instance first) {
for (int m_DatasetKeyColumn : m_DatasetKeyColumns) {
if (first.value(m_DatasetKeyColumn) != m_Template
.value(m_DatasetKeyColumn)) {
return false;
}
}
return true;
}
/**
* Adds the given instance to the dataset
*
* @param inst the instance to add
*/
protected void add(Instance inst) {
m_Dataset.add(inst);
}
/**
* Returns a vector containing the instances in the dataset
*
* @return the current contents
*/
protected ArrayList contents() {
return m_Dataset;
}
/**
* Sorts the instances in the dataset by the run number.
*
* @param runColumn a value of type 'int'
*/
public void sort(int runColumn) {
double[] runNums = new double[m_Dataset.size()];
for (int j = 0; j < runNums.length; j++) {
runNums[j] = m_Dataset.get(j).value(runColumn);
}
int[] index = Utils.stableSort(runNums);
ArrayList newDataset = new ArrayList(runNums.length);
for (int element : index) {
newDataset.add(m_Dataset.get(element));
}
m_Dataset = newDataset;
}
/**
* Returns the revision string.
*
* @return the revision
*/
@Override
public String getRevision() {
return RevisionUtils.extract("$Revision: 11542 $");
}
}
/** Utility class to store the instances in a resultset */
protected class Resultset implements RevisionHandler, Serializable {
/** for serialization. */
private static final long serialVersionUID = 1543786683821339978L;
/** the template */
Instance m_Template;
/** the dataset */
ArrayList m_Datasets;
/**
* Constructir
*
* @param template the template
*/
public Resultset(Instance template) {
m_Template = template;
m_Datasets = new ArrayList();
add(template);
}
/**
* Returns true if the two instances match on those attributes that have
* been designated key columns (eg: scheme name and scheme options)
*
* @param first the first instance
* @return true if first and second match on the currently set key columns
*/
protected boolean matchesTemplate(Instance first) {
for (int m_ResultsetKeyColumn : m_ResultsetKeyColumns) {
if (first.value(m_ResultsetKeyColumn) != m_Template
.value(m_ResultsetKeyColumn)) {
return false;
}
}
return true;
}
/**
* Returns a string descriptive of the resultset key column values for this
* resultset
*
* @return a value of type 'String'
*/
protected String templateString() {
String result = "";
String tempResult = "";
for (int m_ResultsetKeyColumn : m_ResultsetKeyColumns) {
tempResult = m_Template.toString(m_ResultsetKeyColumn) + ' ';
// compact the string
tempResult = Utils.removeSubstring(tempResult, "weka.classifiers.");
tempResult = Utils.removeSubstring(tempResult, "weka.filters.");
tempResult = Utils.removeSubstring(tempResult,
"weka.attributeSelection.");
result += tempResult;
}
return result.trim();
}
/**
* Returns a vector containing all instances belonging to one dataset.
*
* @param inst a template instance
* @return a value of type 'FastVector'
*/
public ArrayList dataset(Instance inst) {
for (int i = 0; i < m_Datasets.size(); i++) {
if (m_Datasets.get(i).matchesTemplate(inst)) {
return m_Datasets.get(i).contents();
}
}
return null;
}
/**
* Adds an instance to this resultset
*
* @param newInst a value of type 'Instance'
*/
public void add(Instance newInst) {
for (int i = 0; i < m_Datasets.size(); i++) {
if (m_Datasets.get(i).matchesTemplate(newInst)) {
m_Datasets.get(i).add(newInst);
return;
}
}
Dataset newDataset = new Dataset(newInst);
m_Datasets.add(newDataset);
}
/**
* Sorts the instances in each dataset by the run number.
*
* @param runColumn a value of type 'int'
*/
public void sort(int runColumn) {
for (int i = 0; i < m_Datasets.size(); i++) {
m_Datasets.get(i).sort(runColumn);
}
}
/**
* Returns the revision string.
*
* @return the revision
*/
@Override
public String getRevision() {
return RevisionUtils.extract("$Revision: 11542 $");
}
} // Resultset
/**
* Returns a string descriptive of the key column values for the "datasets
*
* @param template the template
* @return a value of type 'String'
*/
protected String templateString(Instance template) {
String result = "";
for (int m_DatasetKeyColumn : m_DatasetKeyColumns) {
result += template.toString(m_DatasetKeyColumn) + ' ';
}
if (result.startsWith("weka.classifiers.")) {
result = result.substring("weka.classifiers.".length());
}
return result.trim();
}
/**
* Sets the matrix to use to produce the output.
*
* @param matrix the instance to use to produce the output
* @see ResultMatrix
*/
@Override
public void setResultMatrix(ResultMatrix matrix) {
m_ResultMatrix = matrix;
}
/**
* Gets the instance that produces the output.
*
* @return the instance to produce the output
*/
@Override
public ResultMatrix getResultMatrix() {
return m_ResultMatrix;
}
/**
* Set whether standard deviations are displayed or not.
*
* @param s true if standard deviations are to be displayed
*/
@Override
public void setShowStdDevs(boolean s) {
m_ShowStdDevs = s;
}
/**
* Returns true if standard deviations have been requested.
*
* @return true if standard deviations are to be displayed.
*/
@Override
public boolean getShowStdDevs() {
return m_ShowStdDevs;
}
/**
* Separates the instances into resultsets and by dataset/run.
*
* @throws Exception if the TTest parameters have not been set.
*/
protected void prepareData() throws Exception {
if (m_Instances == null) {
throw new Exception("No instances have been set");
}
if (m_RunColumnSet == -1) {
m_RunColumn = m_Instances.numAttributes() - 1;
} else {
m_RunColumn = m_RunColumnSet;
}
if (m_ResultsetKeyColumnsRange == null) {
throw new Exception("No result specifier columns have been set");
}
m_ResultsetKeyColumnsRange.setUpper(m_Instances.numAttributes() - 1);
m_ResultsetKeyColumns = m_ResultsetKeyColumnsRange.getSelection();
if (m_DatasetKeyColumnsRange == null) {
throw new Exception("No dataset specifier columns have been set");
}
m_DatasetKeyColumnsRange.setUpper(m_Instances.numAttributes() - 1);
m_DatasetKeyColumns = m_DatasetKeyColumnsRange.getSelection();
// Split the data up into result sets
m_Resultsets.clear();
m_DatasetSpecifiers.removeAllSpecifiers();
for (int i = 0; i < m_Instances.numInstances(); i++) {
Instance current = m_Instances.instance(i);
if (current.isMissing(m_RunColumn)) {
throw new Exception("Instance has missing value in run " + "column!\n"
+ current);
}
for (int m_ResultsetKeyColumn : m_ResultsetKeyColumns) {
if (current.isMissing(m_ResultsetKeyColumn)) {
throw new Exception("Instance has missing value in resultset key "
+ "column " + (m_ResultsetKeyColumn + 1) + "!\n" + current);
}
}
for (int m_DatasetKeyColumn : m_DatasetKeyColumns) {
if (current.isMissing(m_DatasetKeyColumn)) {
throw new Exception("Instance has missing value in dataset key "
+ "column " + (m_DatasetKeyColumn + 1) + "!\n" + current);
}
}
boolean found = false;
for (int j = 0; j < m_Resultsets.size(); j++) {
Resultset resultset = m_Resultsets.get(j);
if (resultset.matchesTemplate(current)) {
resultset.add(current);
found = true;
break;
}
}
if (!found) {
Resultset resultset = new Resultset(current);
m_Resultsets.add(resultset);
}
m_DatasetSpecifiers.add(current);
}
// Tell each resultset to sort on the run column
for (int j = 0; j < m_Resultsets.size(); j++) {
Resultset resultset = m_Resultsets.get(j);
if (m_FoldColumn >= 0) {
// sort on folds first in case they are out of order
resultset.sort(m_FoldColumn);
}
resultset.sort(m_RunColumn);
}
m_ResultsetsValid = true;
}
/**
* Gets the number of datasets in the resultsets
*
* @return the number of datasets in the resultsets
*/
@Override
public int getNumDatasets() {
if (!m_ResultsetsValid) {
try {
prepareData();
} catch (Exception ex) {
ex.printStackTrace();
return 0;
}
}
return m_DatasetSpecifiers.numSpecifiers();
}
/**
* Gets the number of resultsets in the data.
*
* @return the number of resultsets in the data
*/
@Override
public int getNumResultsets() {
if (!m_ResultsetsValid) {
try {
prepareData();
} catch (Exception ex) {
ex.printStackTrace();
return 0;
}
}
return m_Resultsets.size();
}
/**
* Gets a string descriptive of the specified resultset.
*
* @param index the index of the resultset
* @return a descriptive string for the resultset
*/
@Override
public String getResultsetName(int index) {
if (!m_ResultsetsValid) {
try {
prepareData();
} catch (Exception ex) {
ex.printStackTrace();
return null;
}
}
return m_Resultsets.get(index).templateString();
}
/**
* Checks whether the resultset with the given index shall be displayed.
*
* @param index the index of the resultset to check whether it shall be
* displayed
* @return whether the specified resultset is displayed
*/
@Override
public boolean displayResultset(int index) {
boolean result;
int i;
result = true;
if (m_DisplayedResultsets != null) {
result = false;
for (i = 0; i < m_DisplayedResultsets.length; i++) {
if (m_DisplayedResultsets[i] == index) {
result = true;
break;
}
}
}
return result;
}
/**
* Computes a paired t-test comparison for a specified dataset between two
* resultsets.
*
* @param datasetSpecifier the dataset specifier
* @param resultset1Index the index of the first resultset
* @param resultset2Index the index of the second resultset
* @param comparisonColumn the column containing values to compare
* @return the results of the paired comparison
* @throws Exception if an error occurs
*/
@Override
public PairedStats calculateStatistics(Instance datasetSpecifier,
int resultset1Index, int resultset2Index, int comparisonColumn)
throws Exception {
if (m_Instances.attribute(comparisonColumn).type() != Attribute.NUMERIC) {
throw new Exception("Comparison column " + (comparisonColumn + 1) + " ("
+ m_Instances.attribute(comparisonColumn).name() + ") is not numeric");
}
if (!m_ResultsetsValid) {
prepareData();
}
Resultset resultset1 = m_Resultsets.get(resultset1Index);
Resultset resultset2 = m_Resultsets.get(resultset2Index);
ArrayList dataset1 = resultset1.dataset(datasetSpecifier);
ArrayList dataset2 = resultset2.dataset(datasetSpecifier);
String datasetName = templateString(datasetSpecifier);
if (dataset1 == null) {
throw new Exception("No results for dataset=" + datasetName
+ " for resultset=" + resultset1.templateString());
} else if (dataset2 == null) {
throw new Exception("No results for dataset=" + datasetName
+ " for resultset=" + resultset2.templateString());
} else if (dataset1.size() != dataset2.size()) {
throw new Exception("Results for dataset=" + datasetName
+ " differ in size for resultset=" + resultset1.templateString()
+ " and resultset=" + resultset2.templateString());
}
PairedStats pairedStats = new PairedStats(m_SignificanceLevel);
for (int k = 0; k < dataset1.size(); k++) {
Instance current1 = dataset1.get(k);
Instance current2 = dataset2.get(k);
if (current1.isMissing(comparisonColumn)) {
System.err.println("Instance has missing value in comparison "
+ "column!\n" + current1);
continue;
}
if (current2.isMissing(comparisonColumn)) {
System.err.println("Instance has missing value in comparison "
+ "column!\n" + current2);
continue;
}
if (current1.value(m_RunColumn) != current2.value(m_RunColumn)) {
System.err.println("Run numbers do not match!\n" + current1 + current2);
}
if (m_FoldColumn != -1) {
if (current1.value(m_FoldColumn) != current2.value(m_FoldColumn)) {
System.err.println("Fold numbers do not match!\n" + current1
+ current2);
}
}
double value1 = current1.value(comparisonColumn);
double value2 = current2.value(comparisonColumn);
pairedStats.add(value1, value2);
}
pairedStats.calculateDerived();
// System.err.println("Differences stats:\n" +
// pairedStats.differencesStats);
return pairedStats;
}
/**
* Creates a key that maps resultset numbers to their descriptions.
*
* @return a value of type 'String'
*/
@Override
public String resultsetKey() {
if (!m_ResultsetsValid) {
try {
prepareData();
} catch (Exception ex) {
ex.printStackTrace();
return ex.getMessage();
}
}
String result = "";
for (int j = 0; j < getNumResultsets(); j++) {
result += "(" + (j + 1) + ") " + getResultsetName(j) + '\n';
}
return result + '\n';
}
/**
* Creates a "header" string describing the current resultsets.
*
* @param comparisonColumn a value of type 'int'
* @return a value of type 'String'
*/
@Override
public String header(int comparisonColumn) {
if (!m_ResultsetsValid) {
try {
prepareData();
} catch (Exception ex) {
ex.printStackTrace();
return ex.getMessage();
}
}
initResultMatrix();
m_ResultMatrix.addHeader("Tester", getClass().getName() + " " + Utils.joinOptions(getOptions()));
m_ResultMatrix.addHeader("Analysing",
m_Instances.attribute(comparisonColumn).name());
m_ResultMatrix.addHeader("Datasets", Integer.toString(getNumDatasets()));
m_ResultMatrix
.addHeader("Resultsets", Integer.toString(getNumResultsets()));
m_ResultMatrix.addHeader("Confidence", getSignificanceLevel()
+ " (two tailed)");
m_ResultMatrix.addHeader("Sorted by", getSortColumnName());
m_ResultMatrix.addHeader("Date",
(new SimpleDateFormat()).format(new Date()));
return m_ResultMatrix.toStringHeader() + "\n";
}
/**
* Carries out a comparison between all resultsets, counting the number of
* datsets where one resultset outperforms the other.
*
* @param comparisonColumn the index of the comparison column
* @param nonSigWin for storing the non-significant wins
* @return a 2d array where element [i][j] is the number of times resultset j
* performed significantly better than resultset i.
* @throws Exception if an error occurs
*/
@Override
public int[][] multiResultsetWins(int comparisonColumn, int[][] nonSigWin)
throws Exception {
int numResultsets = getNumResultsets();
int[][] win = new int[numResultsets][numResultsets];
// int [][] nonSigWin = new int [numResultsets][numResultsets];
for (int i = 0; i < numResultsets; i++) {
for (int j = i + 1; j < numResultsets; j++) {
System.err
.print("Comparing (" + (i + 1) + ") with (" + (j + 1) + ")\r");
System.err.flush();
for (int k = 0; k < getNumDatasets(); k++) {
try {
PairedStats pairedStats = calculateStatistics(
m_DatasetSpecifiers.specifier(k), i, j, comparisonColumn);
if (pairedStats.differencesSignificance < 0) {
win[i][j]++;
} else if (pairedStats.differencesSignificance > 0) {
win[j][i]++;
}
if (pairedStats.differencesStats.mean < 0) {
nonSigWin[i][j]++;
} else if (pairedStats.differencesStats.mean > 0) {
nonSigWin[j][i]++;
}
} catch (Exception ex) {
// ex.printStackTrace();
System.err.println(ex.getMessage());
}
}
}
}
return win;
}
/**
* clears the content and fills the column and row names according to the
* given sorting
*/
protected void initResultMatrix() {
m_ResultMatrix.setSize(getNumResultsets(), getNumDatasets());
m_ResultMatrix.setShowStdDev(m_ShowStdDevs);
for (int i = 0; i < getNumDatasets(); i++) {
m_ResultMatrix.setRowName(i,
templateString(m_DatasetSpecifiers.specifier(i)));
}
for (int j = 0; j < getNumResultsets(); j++) {
m_ResultMatrix.setColName(j, getResultsetName(j));
m_ResultMatrix.setColHidden(j, !displayResultset(j));
}
}
/**
* Carries out a comparison between all resultsets, counting the number of
* datsets where one resultset outperforms the other. The results are
* summarized in a table.
*
* @param comparisonColumn the index of the comparison column
* @return the results in a string
* @throws Exception if an error occurs
*/
@Override
public String multiResultsetSummary(int comparisonColumn) throws Exception {
int[][] nonSigWin = new int[getNumResultsets()][getNumResultsets()];
int[][] win = multiResultsetWins(comparisonColumn, nonSigWin);
initResultMatrix();
m_ResultMatrix.setSummary(nonSigWin, win);
return m_ResultMatrix.toStringSummary();
}
/**
* returns a ranking of the resultsets
*
* @param comparisonColumn the column to compare with
* @return the ranking
* @throws Exception if something goes wrong
*/
@Override
public String multiResultsetRanking(int comparisonColumn) throws Exception {
int[][] nonSigWin = new int[getNumResultsets()][getNumResultsets()];
int[][] win = multiResultsetWins(comparisonColumn, nonSigWin);
initResultMatrix();
m_ResultMatrix.setRanking(win);
return m_ResultMatrix.toStringRanking();
}
/**
* Creates a comparison table where a base resultset is compared to the other
* resultsets. Results are presented for every dataset.
*
* @param baseResultset the index of the base resultset
* @param comparisonColumn the index of the column to compare over
* @return the comparison table string
* @throws Exception if an error occurs
*/
@Override
public String multiResultsetFull(int baseResultset, int comparisonColumn)
throws Exception {
int maxWidthMean = 2;
int maxWidthStdDev = 2;
double[] sortValues = new double[getNumDatasets()];
// determine max field width
for (int i = 0; i < getNumDatasets(); i++) {
sortValues[i] = Double.POSITIVE_INFINITY; // sorts skipped cols to end
for (int j = 0; j < getNumResultsets(); j++) {
if (!displayResultset(j)) {
continue;
}
try {
PairedStats pairedStats = calculateStatistics(
m_DatasetSpecifiers.specifier(i), baseResultset, j,
comparisonColumn);
if (!Double.isInfinite(pairedStats.yStats.mean)
&& !Double.isNaN(pairedStats.yStats.mean)) {
double width = ((Math.log(Math.abs(pairedStats.yStats.mean)) / Math
.log(10)) + 1);
if (width > maxWidthMean) {
maxWidthMean = (int) width;
}
}
if (j == baseResultset) {
if (getSortColumn() != -1) {
sortValues[i] = calculateStatistics(
m_DatasetSpecifiers.specifier(i), baseResultset, j,
getSortColumn()).xStats.mean;
} else {
sortValues[i] = i;
}
}
if (m_ShowStdDevs && !Double.isInfinite(pairedStats.yStats.stdDev)
&& !Double.isNaN(pairedStats.yStats.stdDev)) {
double width = ((Math.log(Math.abs(pairedStats.yStats.stdDev)) / Math
.log(10)) + 1);
if (width > maxWidthStdDev) {
maxWidthStdDev = (int) width;
}
}
} catch (Exception ex) {
// ex.printStackTrace();
System.err.println(ex);
}
}
}
// sort rows according to sort column
m_SortOrder = Utils.sort(sortValues);
// determine column order
m_ColOrder = new int[getNumResultsets()];
m_ColOrder[0] = baseResultset;
int index = 1;
for (int i = 0; i < getNumResultsets(); i++) {
if (i == baseResultset) {
continue;
}
m_ColOrder[index] = i;
index++;
}
// setup matrix
initResultMatrix();
m_ResultMatrix.setRowOrder(m_SortOrder);
m_ResultMatrix.setColOrder(m_ColOrder);
m_ResultMatrix.setMeanWidth(maxWidthMean);
m_ResultMatrix.setStdDevWidth(maxWidthStdDev);
m_ResultMatrix.setSignificanceWidth(1);
// make sure that test base is displayed, even though it might not be
// selected
for (int i = 0; i < m_ResultMatrix.getColCount(); i++) {
if ((i == baseResultset) && (m_ResultMatrix.getColHidden(i))) {
m_ResultMatrix.setColHidden(i, false);
System.err.println("Note: test base was hidden - set visible!");
}
}
// the data
for (int i = 0; i < getNumDatasets(); i++) {
m_ResultMatrix.setRowName(i,
templateString(m_DatasetSpecifiers.specifier(i)));
for (int j = 0; j < getNumResultsets(); j++) {
try {
// calc stats
PairedStats pairedStats = calculateStatistics(
m_DatasetSpecifiers.specifier(i), baseResultset, j,
comparisonColumn);
// count
m_ResultMatrix.setCount(i, pairedStats.count);
// mean
m_ResultMatrix.setMean(j, i, pairedStats.yStats.mean);
// std dev
m_ResultMatrix.setStdDev(j, i, pairedStats.yStats.stdDev);
// significance
if (pairedStats.differencesSignificance < 0) {
m_ResultMatrix.setSignificance(j, i, ResultMatrix.SIGNIFICANCE_WIN);
} else if (pairedStats.differencesSignificance > 0) {
m_ResultMatrix
.setSignificance(j, i, ResultMatrix.SIGNIFICANCE_LOSS);
} else {
m_ResultMatrix.setSignificance(j, i, ResultMatrix.SIGNIFICANCE_TIE);
}
} catch (Exception e) {
// e.printStackTrace();
System.err.println(e);
}
}
}
// generate output
StringBuffer result = new StringBuffer(1000);
try {
result.append(m_ResultMatrix.toStringMatrix());
} catch (Exception e) {
e.printStackTrace();
}
// append a key so that we can tell the difference between long
// scheme+option names
if (m_ResultMatrix.getEnumerateColNames()) {
result.append("\n\n" + m_ResultMatrix.toStringKey());
}
return result.toString();
}
/**
* Lists options understood by this object.
*
* @return an enumeration of Options.
*/
@Override
public Enumeration © 2015 - 2024 Weber Informatics LLC | Privacy Policy