Changeset 5329
- Timestamp:
- 11/01/04 19:10:39 (20 years ago)
- Location:
- trunk/MagicSoft/Mars
- Files:
-
- 2 edited
Legend:
- Unmodified
- Added
- Removed
-
trunk/MagicSoft/Mars/Changelog
r5328 r5329 20 20 21 21 -*-*- END OF LINE -*-*- 22 23 2004/11/01: David Paneque 24 25 * mtemp/mmpi/SupercutsONOFFClasses/MHFindSignificanceONOFF.cc 26 27 - Condition to assume a constant background in the fit 28 of the ON and OFF data is softened: 29 1) Bins with zero content are allowed (if they are less than 10% 30 of the total number of bins used). Yet their error is set to 31 a huge number 10E19; which means that they are not taken into account 32 in the fit. 33 2) Maximum number of bins with low content (less than 9) is increased 34 from 5% to 20%. The error of these bins is set to 3. 35 36 The motivation of these changes is to allow the fitting procedures also 37 in those cases where the background rejection is that large that very few 38 events remain in the background region of the alpha plot. This situation 39 is occuring quite often when using the Random Forest and SPECIALLY, when 40 making large SIZE cuts. 41 42 43 22 44 2004/10/30: Markus Gaug 23 45 … … 25 47 * msignal/MExtractTime.h 26 48 - set the version number by 1 higher due to the changes on 12.10. 49 27 50 28 51 -
trunk/MagicSoft/Mars/mtemp/mmpi/SupercutsONOFFClasses/MHFindSignificanceONOFF.cc
r4411 r5329 1047 1047 // count bins with zero entry 1048 1048 if (content <= 0.0) 1049 fNzeroOFF++; 1050 1049 { 1050 fNzeroOFF++; 1051 // The error of the bin is set to a huge number, 1052 // so that it does not have any weight in the fit 1053 fHistOFF->SetBinError(i, dummy); 1054 } 1051 1055 // set minimum error 1052 1056 if (content < 9.0) … … 1151 1155 1152 1156 fConstantBackg = kFALSE; 1153 if ( fNzeroOFF > 0 || (Double_t)fMlowOFF>0.05*(Double_t)fMbinsOFF ) 1157 1158 // Condition for disabling the fitting procedure and 1159 // assuming a constant background (before Nov 2004) 1160 1161 // if ( fNzeroOFF > 0 || (Double_t)fMlowOFF>0.05*(Double_t)fMbinsOFF ) 1162 1163 1164 // Condition for disabling the fitting procedure and 1165 // assuming a constant background (After Nov 01 2004) 1166 // I softened the condition to allow the fit also in situations 1167 // where the reduction of the background is such that very 1168 // few events survived; which is 1169 // Specially frequent with Random Forest at high Sizes) 1170 1171 if ( (Double_t)fNzeroOFF > 0.1*(Double_t)fMbinsOFF || 1172 (Double_t)fMlowOFF > 0.2*(Double_t)fMbinsOFF ) 1154 1173 { 1155 1174 *fLog << "MHFindSignificanceONOFF::FitPolynomialOFF; polynomial fit not possible, fNzeroOFF, fMlowOFF, fMbinsOFF = " … … 1182 1201 Double_t val, err; 1183 1202 val = mean; 1184 err = sqrt( mean / (Double_t)fMbinsOFF );1203 err = rms; // sqrt( mean / (Double_t)fMbinsOFF ); 1185 1204 1186 1205 fPolyOFF->SetParameter(0, val); … … 1553 1572 // count bins with zero entry 1554 1573 if (content <= 0.0) 1555 fNzero++; 1556 1574 { 1575 fNzero++; 1576 // The error of the bin is set to a huge number, 1577 // so that it does not have any weight in the fit 1578 fHistOFF->SetBinError(i, dummy); 1579 } 1580 1557 1581 // set minimum error 1558 1582 if (content < 9.0) … … 1657 1681 1658 1682 fConstantBackg = kFALSE; 1659 if ( fNzero > 0 || (Double_t)fMlow>0.05*(Double_t)fMbins ) 1683 1684 // Condition for disabling the fitting procedure and 1685 // assuming a constant background (before Nov 2004) 1686 1687 // if ( fNzero > 0 || (Double_t)fMlow>0.05*(Double_t)fMbins ) 1688 1689 1690 // Condition for disabling the fitting procedure and 1691 // assuming a constant background (After Nov 01 2004) 1692 // I softened the condition to allow the fit also in situations 1693 // where the reduction of the background is such that very 1694 // few events survived; which is 1695 // Specially frequent with Random Forest at high Sizes) 1696 1697 if ( (Double_t)fNzero > 0.1*(Double_t)fMbins || 1698 (Double_t)fMlow > 0.2*(Double_t)fMbins ) 1699 1660 1700 { 1661 1701 *fLog << "MHFindSignificanceONOFF::FitPolynomial; polynomial fit not possible, fNzero, fMlow, fMbins = " … … 1688 1728 Double_t val, err; 1689 1729 val = mean; 1690 err = sqrt( mean / (Double_t)fMbins );1730 err = rms; // sqrt( mean / (Double_t)fMbins ); 1691 1731 1692 1732 fPoly->SetParameter(0, val);
Note:
See TracChangeset
for help on using the changeset viewer.