public void testNaNs() {
SimpleRegression regression = new SimpleRegression();
assertTrue("intercept not NaN", Double.isNaN(regression.getIntercept()));
assertTrue("slope not NaN", Double.isNaN(regression.getSlope()));
assertTrue("slope std err not NaN", Double.isNaN(regression.getSlopeStdErr()));
assertTrue("intercept std err not NaN", Double.isNaN(regression.getInterceptStdErr()));
assertTrue("MSE not NaN", Double.isNaN(regression.getMeanSquareError()));
assertTrue("e not NaN", Double.isNaN(regression.getR()));
assertTrue("r-square not NaN", Double.isNaN(regression.getRSquare()));
assertTrue("RSS not NaN", Double.isNaN(regression.getRegressionSumSquares()));
assertTrue("SSE not NaN", Double.isNaN(regression.getSumSquaredErrors()));
assertTrue("SSTO not NaN", Double.isNaN(regression.getTotalSumSquares()));
assertTrue("predict not NaN", Double.isNaN(regression.predict(0)));
regression.addData(1, 2);
regression.addData(1, 3);
// No x variation, so these should still blow...
assertTrue("intercept not NaN", Double.isNaN(regression.getIntercept()));
assertTrue("slope not NaN", Double.isNaN(regression.getSlope()));
assertTrue("slope std err not NaN", Double.isNaN(regression.getSlopeStdErr()));
assertTrue("intercept std err not NaN", Double.isNaN(regression.getInterceptStdErr()));
assertTrue("MSE not NaN", Double.isNaN(regression.getMeanSquareError()));
assertTrue("e not NaN", Double.isNaN(regression.getR()));
assertTrue("r-square not NaN", Double.isNaN(regression.getRSquare()));
assertTrue("RSS not NaN", Double.isNaN(regression.getRegressionSumSquares()));
assertTrue("SSE not NaN", Double.isNaN(regression.getSumSquaredErrors()));
assertTrue("predict not NaN", Double.isNaN(regression.predict(0)));
// but SSTO should be OK
assertTrue("SSTO NaN", !Double.isNaN(regression.getTotalSumSquares()));
regression = new SimpleRegression();
regression.addData(1, 2);
regression.addData(3, 3);
// All should be OK except MSE, s(b0), s(b1) which need one more df
assertTrue("interceptNaN", !Double.isNaN(regression.getIntercept()));
assertTrue("slope NaN", !Double.isNaN(regression.getSlope()));
assertTrue("slope std err not NaN", Double.isNaN(regression.getSlopeStdErr()));
assertTrue("intercept std err not NaN", Double.isNaN(regression.getInterceptStdErr()));
assertTrue("MSE not NaN", Double.isNaN(regression.getMeanSquareError()));
assertTrue("r NaN", !Double.isNaN(regression.getR()));
assertTrue("r-square NaN", !Double.isNaN(regression.getRSquare()));
assertTrue("RSS NaN", !Double.isNaN(regression.getRegressionSumSquares()));
assertTrue("SSE NaN", !Double.isNaN(regression.getSumSquaredErrors()));
assertTrue("SSTO NaN", !Double.isNaN(regression.getTotalSumSquares()));
assertTrue("predict NaN", !Double.isNaN(regression.predict(0)));
regression.addData(1, 4);
// MSE, MSE, s(b0), s(b1) should all be OK now
assertTrue("MSE NaN", !Double.isNaN(regression.getMeanSquareError()));
assertTrue("slope std err NaN", !Double.isNaN(regression.getSlopeStdErr()));
assertTrue("intercept std err NaN", !Double.isNaN(regression.getInterceptStdErr()));
}
public void testClear() {
SimpleRegression regression = new SimpleRegression();
regression.addData(corrData);
assertEquals("number of observations", 17, regression.getN());
regression.clear();
assertEquals("number of observations", 0, regression.getN());
regression.addData(corrData);
assertEquals("r-square", .896123, regression.getRSquare(), 10E-6);
regression.addData(data);
assertEquals("number of observations", 53, regression.getN());
}
/**
* Adds a sample to the model with the given frequency and length values.
*
* @param averageStepFrequency
* @param averageStepLength
*/
public void addSampleToModel(Double averageStepFrequency, Double averageStepLength) {
// ...insert it into the regression model
reg.addData(averageStepFrequency, averageStepLength);
// ...and sample list
sampleList.add(new AutoGaitSample(averageStepFrequency, averageStepLength));
}
public AutoGaitModel(double[][] samples) {
// Add samples to the regression model
reg.addData(samples);
// Add values to the sample list
for (double[] line : samples) addSampleToModel(line[0], line[1]);
}
// Test remove multiple observations
public void testRemoveMultiple() throws Exception {
// Create regression with inference data then remove to test
SimpleRegression regression = new SimpleRegression();
regression.addData(infData);
regression.removeData(removeMultiple);
regression.addData(removeMultiple);
// Use the inference assertions to make sure that everything worked
assertEquals("slope std err", 0.011448491, regression.getSlopeStdErr(), 1E-10);
assertEquals("std err intercept", 0.286036932, regression.getInterceptStdErr(), 1E-8);
assertEquals("significance", 4.596e-07, regression.getSignificance(), 1E-8);
assertEquals(
"slope conf interval half-width",
0.0270713794287,
regression.getSlopeConfidenceInterval(),
1E-8);
}
public void testCorr() {
SimpleRegression regression = new SimpleRegression();
regression.addData(corrData);
assertEquals("number of observations", 17, regression.getN());
assertEquals("r-square", .896123, regression.getRSquare(), 10E-6);
assertEquals("r", -0.94663767742, regression.getR(), 1E-10);
}
// Jira MATH-85 = Bugzilla 39432
public void testSSENonNegative() {
double[] y = {8915.102, 8919.302, 8923.502};
double[] x = {1.107178495E2, 1.107264895E2, 1.107351295E2};
SimpleRegression reg = new SimpleRegression();
for (int i = 0; i < x.length; i++) {
reg.addData(x[i], y[i]);
}
assertTrue(reg.getSumSquaredErrors() >= 0.0);
}
public void testRandom() throws Exception {
SimpleRegression regression = new SimpleRegression();
Random random = new Random(1);
int n = 100;
for (int i = 0; i < n; i++) {
regression.addData(((double) i) / (n - 1), random.nextDouble());
}
assertTrue(0.0 < regression.getSignificance() && regression.getSignificance() < 1.0);
}
public void testPerfectNegative() throws Exception {
SimpleRegression regression = new SimpleRegression();
int n = 100;
for (int i = 0; i < n; i++) {
regression.addData(-((double) i) / (n - 1), i);
}
assertEquals(0.0, regression.getSignificance(), 1.0e-5);
assertTrue(regression.getSlope() < 0.0);
}
public void testInference() throws Exception {
// ---------- verified against R, version 1.8.1 -----
// infData
SimpleRegression regression = new SimpleRegression();
regression.addData(infData);
assertEquals("slope std err", 0.011448491, regression.getSlopeStdErr(), 1E-10);
assertEquals("std err intercept", 0.286036932, regression.getInterceptStdErr(), 1E-8);
assertEquals("significance", 4.596e-07, regression.getSignificance(), 1E-8);
assertEquals(
"slope conf interval half-width",
0.0270713794287,
regression.getSlopeConfidenceInterval(),
1E-8);
// infData2
regression = new SimpleRegression();
regression.addData(infData2);
assertEquals("slope std err", 1.07260253, regression.getSlopeStdErr(), 1E-8);
assertEquals("std err intercept", 4.17718672, regression.getInterceptStdErr(), 1E-8);
assertEquals("significance", 0.261829133982, regression.getSignificance(), 1E-11);
assertEquals(
"slope conf interval half-width",
2.97802204827,
regression.getSlopeConfidenceInterval(),
1E-8);
// ------------- End R-verified tests -------------------------------
// FIXME: get a real example to test against with alpha = .01
assertTrue(
"tighter means wider",
regression.getSlopeConfidenceInterval() < regression.getSlopeConfidenceInterval(0.01));
try {
regression.getSlopeConfidenceInterval(1);
fail("expecting IllegalArgumentException for alpha = 1");
} catch (IllegalArgumentException ex) {
// ignored
}
}
/**
* Computes the Pearson's product-moment correlation coefficient between the two arrays. Throws
* IllegalArgumentException if the arrays do not have the same length or their common length is
* less than 2
*
* @param xArray first data array
* @param yArray second data array
* @return Returns Pearson's correlation coefficient for the two arrays
* @throws IllegalArgumentException if the arrays lengths do not match or there is insufficient
* data
*/
public double correlation(final double[] xArray, final double[] yArray)
throws IllegalArgumentException {
SimpleRegression regression = new SimpleRegression();
if (xArray.length == yArray.length && xArray.length > 1) {
for (int i = 0; i < xArray.length; i++) {
regression.addData(xArray[i], yArray[i]);
}
return regression.getR();
} else {
throw MathRuntimeException.createIllegalArgumentException(
"invalid array dimensions. xArray has size {0}; yArray has {1} elements",
xArray.length, yArray.length);
}
}
public void testNorris() {
SimpleRegression regression = new SimpleRegression();
for (int i = 0; i < data.length; i++) {
regression.addData(data[i][1], data[i][0]);
}
// Tests against certified values from
// http://www.itl.nist.gov/div898/strd/lls/data/LINKS/DATA/Norris.dat
assertEquals("slope", 1.00211681802045, regression.getSlope(), 10E-12);
assertEquals("slope std err", 0.429796848199937E-03, regression.getSlopeStdErr(), 10E-12);
assertEquals("number of observations", 36, regression.getN());
assertEquals("intercept", -0.262323073774029, regression.getIntercept(), 10E-12);
assertEquals("std err intercept", 0.232818234301152, regression.getInterceptStdErr(), 10E-12);
assertEquals("r-square", 0.999993745883712, regression.getRSquare(), 10E-12);
assertEquals("SSR", 4255954.13232369, regression.getRegressionSumSquares(), 10E-9);
assertEquals("MSE", 0.782864662630069, regression.getMeanSquareError(), 10E-10);
assertEquals("SSE", 26.6173985294224, regression.getSumSquaredErrors(), 10E-9);
// ------------ End certified data tests
assertEquals("predict(0)", -0.262323073774029, regression.predict(0), 10E-12);
assertEquals("predict(1)", 1.00211681802045 - 0.262323073774029, regression.predict(1), 10E-12);
}
/**
* Adds the observations represented by the elements in <code>data</code>.
*
* <p><code>(data[0][0],data[0][1])</code> will be the first observation, then <code>
* (data[1][0],data[1][1])</code>, etc.
*
* <p>This method does not replace data that has already been added. The observations represented
* by <code>data</code> are added to the existing dataset.
*
* <p>To replace all data, use <code>clear()</code> before adding the new data.
*
* @param data array of observations to be added
*/
public void addData(double[][] data) {
for (int i = 0; i < data.length; i++) {
addData(data[i][0], data[i][1]);
}
}
// Remove multiple observations past empty (i.e. size of array > n)
public void testRemoveMultiplePastEmpty() {
SimpleRegression regression = new SimpleRegression();
regression.addData(removeX, removeY);
regression.removeData(removeMultiple);
assertEquals(regression.getN(), 0);
}
// Remove single observation to empty
public void testRemoveObsFromSingle() {
SimpleRegression regression = new SimpleRegression();
regression.addData(removeX, removeY);
regression.removeData(removeX, removeY);
assertEquals(regression.getN(), 0);
}
public void SummuarySimulation() {
calculatePower();
PrintStream Pout1 = null;
PrintStream Pout2 = null;
PrintStream Pout3 = null;
PrintStream Pout4 = null;
PrintStream Pout5 = null;
PrintStream Pout6 = null;
ChiSquaredDistribution chi = new ChiSquaredDistributionImpl(weight.length);
try {
Pout1 = new PrintStream(new BufferedOutputStream(new FileOutputStream("LogAP.txt")));
Pout2 = new PrintStream(new BufferedOutputStream(new FileOutputStream("LNP.txt")));
Pout3 = new PrintStream(new BufferedOutputStream(new FileOutputStream("LN.txt")));
Pout4 = new PrintStream(new BufferedOutputStream(new FileOutputStream("Wald.txt")));
Pout5 = new PrintStream(new BufferedOutputStream(new FileOutputStream("F.txt")));
Pout6 = new PrintStream(new BufferedOutputStream(new FileOutputStream("LogDP.txt")));
} catch (Exception E) {
E.printStackTrace(System.err);
}
for (int i = 0; i < SimulationResults.size(); i++) {
ArrayList PointStatistics = (ArrayList) SimulationResults.get(i);
double[] LRT = new double[PointStatistics.size()];
double[] WALD = new double[PointStatistics.size()];
DescriptiveStatistics dsLRT = new DescriptiveStatisticsImpl();
DescriptiveStatistics dsWALD = new DescriptiveStatisticsImpl();
SimpleRegression sr = new SimpleRegression();
for (int j = 0; j < PointStatistics.size(); j++) {
PointMappingStatistic pms = (PointMappingStatistic) PointStatistics.get(j);
double lod = pms.get_LOD() > 0 ? pms.get_LOD() : 0;
double ln = pms.get_LN();
double p = 0;
try {
p = chi.cumulativeProbability(ln);
} catch (Exception E) {
E.printStackTrace(System.err);
}
double logLOD = -1 * Math.log10(1 - p);
Pout1.print(pms.get_logP_additive() + " ");
Pout2.print(logLOD + " ");
Pout3.print(pms.get_LN() + " ");
Pout4.print(pms.get_wald() + " ");
Pout5.print(pms.get_logP_F() + " ");
Pout6.print(pms.get_logP_dominance() + " ");
dsLRT.addValue(pms.get_LN());
dsWALD.addValue(pms.get_wald());
LRT[j] = pms.get_LN();
WALD[j] = pms.get_wald();
sr.addData(pms.get_LN(), pms.get_wald());
}
System.out.println(
dsLRT.getMean()
+ " +- "
+ dsLRT.getStandardDeviation()
+ " "
+ dsWALD.getMean()
+ " +- "
+ dsWALD.getStandardDeviation()
+ " cor "
+ sr.getR());
dsLRT.clear();
dsWALD.clear();
sr.clear();
Pout1.println();
Pout2.println();
Pout3.println();
Pout4.println();
Pout5.println();
Pout6.println();
}
Pout1.close();
Pout2.close();
Pout3.close();
Pout4.close();
Pout5.close();
Pout6.close();
}