Permalink
Cannot retrieve contributors at this time
Fetching contributors…
Cannot retrieve contributors at this time
| /* PDAP:PDTREE package for Mesquite copyright 2001-2009 P. Midford & W. MaddisonPDAP:PDTREE is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY.The web site for PDAP:PDTREE is http://mesquiteproject.org/pdap_mesquite/This source code and its compiled class files are free and modifiable under the terms of GNU Lesser General Public License. (http://www.gnu.org/copyleft/lesser.html) */package mesquite.pdap.SPRegDiagnostics;/*~~ */import mesquite.lib.*;import mesquite.pdap.lib.*;import JSci.maths.statistics.*; // for p-value calculations/** * @author Peter E. Midford */public class SPRegDiagnostics extends StatPakSource { public boolean startJob(String arguments, Object condition, boolean hiredByName) { return true; } public PDAPStatPak getStatPak(){ return new regStatPak(); } public String getName(){ return "Regression Diagnostics StatPak Source"; } /*.................................................................................................................*/ public String getAuthors() { return "Peter E. Midford, Ted Garland Jr., and Wayne P. Maddison" ; } /*.................................................................................................................*/ /** returns module version */ public String getVersion() { return "1.15"; } /*.................................................................................................................*/ public boolean isPrerelease() { return false; } /*.................................................................................................................*/ public boolean isSubstantive() { return true; // the statPak is substantive, but I don't think this module is... } public String getExplanation() { return "Calculates summary statistics and regression coefficients for a scatterplot."; }}/** * This statPak is for standard (not through the origin or based on residuals) regressions * @author Peter E. Midford * created 2002 * */class regStatPak extends PDAPStatPak { double r2 = MesquiteDouble.unassigned; //r squared value for least squares regression double ts = MesquiteDouble.unassigned; //T statistic double f = MesquiteDouble.unassigned; //F statistic double pvalue = MesquiteDouble.unassigned; public regStatPak(){ super(); } /** * @param xValues values for first trait * @param yValues values for second trait * @param chart the boundaries of the chart are reset * @return false */ public boolean doCalculations(NumberArray xValues, NumberArray yValues, MesquiteChart chart){ // here you can put the main calculations. resetChartBounds(chart); MesquiteDouble sx = new MesquiteDouble(0.0); MesquiteDouble sx2 = new MesquiteDouble(0.0); MesquiteDouble sy = new MesquiteDouble(0.0); MesquiteDouble sxy = new MesquiteDouble(0.0); mean[0] = 0; mean[1] = 0; variance[0] = 0; variance[1] = 0; corr = 0; calcBounds(xValues,yValues); getMeans(xValues,yValues); mean[0] /= getNumber(); mean[1] /= getNumber(); getVars(xValues,yValues); variance[0] /= getNumberMinus1(); variance[1] /= getNumberMinus1(); corr /= getNumberMinus1(); cov = corr; if ((variance[0] > 0) && (variance[1] > 0)) corr = corr/Math.sqrt(variance[0]*variance[1]); else corr = 0; getSums(xValues,yValues,sx,sx2,sy,sxy); if ((getNumber()*sx2.getValue()-sx.getValue()*sx.getValue()) != 0) { lsregok = true; double d = 1/(getNumber()*sx2.getValue()-sx.getValue()*sx.getValue()); ls = d*(getNumber()*sxy.getValue()-sx.getValue()*sy.getValue()); lsy = d*(sx2.getValue()*sy.getValue()-sx.getValue()*sxy.getValue()); } else lsregok = false; finishStat(); if (Math.abs(corr) <1 && df > 0) if (ts<0) pvalue= tStats.cumulative(ts); else pvalue = 1-tStats.cumulative(ts); else pvalue = MesquiteDouble.unassigned; return false; } /** * @return p-value saved in the statPak */ public double getPValue(){ return pvalue; } private void getMeans(NumberArray data0, NumberArray data1){ for (int i = 0; i < data0.getSize(); i++) if (data0.isCombinable(i) && data1.isCombinable(i)) { mean[0] += data0.getDouble(i); mean[1] += data1.getDouble(i); } } private void getVars(NumberArray data0, NumberArray data1){ for (int i = 0; i < data0.getSize(); i++) if (data0.isCombinable(i) && data1.isCombinable(i)) { double dev0 = mean[0] - data0.getDouble(i); double dev1 = mean[1] - data1.getDouble(i); variance[0] += dev0*dev0; variance[1] += dev1*dev1; corr += dev0*dev1; } } // This calculates the rma and the y intercept for rma called rmay, ma and // the y intercept for the ma called may, r squared called r2, t statistic // called ts, the f statistic, and the degrees of freedom called df. private void finishStat() { if (variance[0]>0) { rma = Math.sqrt(variance[1]/variance[0]); if (corr<0) // make sure the sign of the reduced major axis slope agrees with rma *= -1; // that of the correlation rmay = mean[1]-rma*mean[0]; } double d = Math.sqrt(((variance[0]+variance[1])*(variance[0]+variance[1])) - 4*(variance[0]*variance[1]-cov*cov)); double lambda1 = (variance[0]+variance[1]+d)/2; // Sokal and Rolf 1981 set Y1 for vertical axis, for example on 595-596. if (lambda1 == variance[1]) { if (cov >0) ma = MesquiteDouble.infinite; else ma = MesquiteDouble.negInfinite; } else ma = cov/(lambda1-variance[1]); // This is correct variance[1] is vertical axis may = mean[1]-ma*mean[0]; r2 = corr*corr; if (r2<1.0) { ts = corr*Math.sqrt((getNumber()-2)/(1-r2)); f = ts*ts; } df = getNumber()-2; // nominal guess if (df > 0) tStats = new TDistribution(df); // have a df estimate, so now is time to make tStats } /*............................................................*/ /** * This reports a statistical summary from the statPak * * @return string containing summary statistics */ public String flst() { if (!MesquiteDouble.isCombinable(ls)) return ""; StringBuffer results = new StringBuffer(400); results.append("Summary statistics for regression on scatterplot.\n\n"); results.append("Number of data points: " + getNumber() + "\n"); results.append("Minimum of X,Y coordinates: " + min[0] + ", " + min[1] + "\n"); results.append("Maximum of X,Y coordinates: " + max[0] + ", " + max[1] + "\n"); results.append("Mean of X,Y coordinates: " + mean[0] + ", " + mean[1] + "\n"); results.append("Variance of X,Y coordinates: " + variance[0] + ", " + variance[1] + "\n"); results.append("Covariance: " + cov + "\n"); results.append("Pearson Product-Moment Correlation Coefficient: " + corr + "\n"); results.append("Reduced Major Axis.\n"); if (lsregok) { results.append(" Slope: " + rma + "\n"); results.append(" Y Intercept: " + rmay + "\n"); } else { results.append(" Slope: undefined \n"); results.append(" Y Intercept: undefined \n"); } results.append("Major Axis.\n"); results.append(" Slope: " + ma + "\n"); results.append(" Y Intercept: " + may + "\n"); results.append("Least Squares Regression.\n"); if (lsregok) { results.append(" Slope: " + ls + "\n"); results.append(" Intercept: " + lsy + "\n"); results.append(" R Squared: " + r2 + "\n"); } else { results.append(" Slope: undefined\n"); results.append(" Intercept: undefined\n"); results.append(" R Squared: undefined\n"); } if (Math.abs(corr) <1) { results.append("t: "+ ts + "\n"); results.append("F: " + f + "\n"); results.append("d.f.: " + df + "\n"); if (df > 0) { double tail1; if (ts<0) tail1= tStats.cumulative(ts); else tail1 = 1-tStats.cumulative(ts); results.append("p-value:\n"); results.append(" 2-tailed: " + (2.0*tail1) + "\n"); results.append(" 1-tailed: " + tail1 + "\n"); } } else { results.append("t: undefined\n"); results.append("F: undefined\n"); } results.append("\n\n"); return results.toString(); } /** * @return set of numbers with captions summarizing the contents of the statPak */ public MesquiteNumber[] getWritableResults() { if (writableResults == null){ writableResults = new MesquiteNumber[23]; writableResults[0] = new MesquiteNumber(); writableResults[0].setName("Number of contrasts"); writableResults[1] = new MesquiteNumber(); writableResults[1].setName("Min X"); writableResults[2] = new MesquiteNumber(); writableResults[2].setName("Min Y"); writableResults[3] = new MesquiteNumber(); writableResults[3].setName("Max X"); writableResults[4] = new MesquiteNumber(); writableResults[4].setName("Max Y"); writableResults[5] = new MesquiteNumber(); writableResults[5].setName("Mean X"); writableResults[6] = new MesquiteNumber(); writableResults[6].setName("Mean Y"); writableResults[7] = new MesquiteNumber(); writableResults[7].setName("Variance X"); writableResults[8] = new MesquiteNumber(); writableResults[8].setName("Variance Y" ); writableResults[9] = new MesquiteNumber(); writableResults[9].setName("Covariance"); writableResults[10] = new MesquiteNumber(); writableResults[10].setName("Pearson Product-Moment Correlation Coefficient"); writableResults[11] = new MesquiteNumber(); writableResults[11].setName("RMA Slope"); writableResults[12] = new MesquiteNumber(); writableResults[12].setName("RMA Intercept"); writableResults[13] = new MesquiteNumber(); writableResults[13].setName("MA Slope"); writableResults[14] = new MesquiteNumber(); writableResults[14].setName("MA Intercept"); writableResults[15] = new MesquiteNumber(); writableResults[15].setName("OLS Slope"); writableResults[16] = new MesquiteNumber(); writableResults[16].setName("OLS Intercept"); writableResults[17] = new MesquiteNumber(); writableResults[17].setName("r-squared"); writableResults[18] = new MesquiteNumber(); writableResults[18].setName("t"); writableResults[19] = new MesquiteNumber(); writableResults[19].setName("F"); writableResults[20] = new MesquiteNumber(); writableResults[20].setName("df"); writableResults[21] = new MesquiteNumber(); writableResults[21].setName("p (2-tailed)"); writableResults[22] = new MesquiteNumber(); writableResults[22].setName("p (1-tailed)"); } if (MesquiteDouble.isCombinable(corr)){ writableResults[0].setValue(getNumber()); writableResults[1].setValue(min[0]); writableResults[2].setValue(min[1]); writableResults[3].setValue(max[0]); writableResults[4].setValue(max[1]); writableResults[5].setValue(mean[0]); writableResults[6].setValue(mean[1]); writableResults[7].setValue(variance[0]); writableResults[8].setValue(variance[1]); writableResults[9].setValue(cov); writableResults[10].setValue(corr); writableResults[11].setValue(rma); writableResults[12].setValue(rmay); writableResults[13].setValue(ma); writableResults[14].setValue(may); writableResults[15].setValue(ls); writableResults[16].setValue(lsy); writableResults[17].setValue(r2); writableResults[18].setValue(ts); writableResults[19].setValue(f); writableResults[20].setValue(df); if (df > 0) { double tail1; if (ts<0) tail1= tStats.cumulative(ts); else tail1 = 1-tStats.cumulative(ts); writableResults[21].setValue(2*tail1); writableResults[22].setValue(tail1); } else{ writableResults[21].setToUnassigned(); writableResults[22].setToUnassigned(); } } else for (int i=0;i<writableResults.length;i++) writableResults[i].setToUnassigned(); return writableResults; } /*---------------------*/ /** * Support for Mesquite 2.5+ 'step through trees' */ public String getLegendText() { if (!MesquiteDouble.isCombinable(ts)){ return ""; } StringBuffer results = new StringBuffer(200); results.append("Number of data points: " + getNumber() + "\n"); results.append("Pearson Product-Moment Correlation Coefficient: " + corr + "\n"); if (Math.abs(corr) <1) { if (df > 0) results.append("Two tailed p-value: " + (2.0*(1-tStats.cumulative(Math.abs(ts)))) + "\n"); } return results.toString(); }} //class regStatPak |