1. 1/16 MDC post-mortem redux Status as of last CC meeting: –True values of cross-section and oscillation parameters were used to reweight the ND and FD MC samples to perform a match-up with MDC datasets. No obvious improvement as agreement was OK even with nominal systematic parameters –BMPT parameters unknown. Used -7% normalisation as a guesstimate of the effect of BMPT reweighting. This value did not appear to be supported by fits to the MDC data. This meeting –Obtained BMPT reweighting histograms that were generated by Mark for the MDC –Used BMPTHistoWeightCalculator method in MCReweight to calculate a weighting factor for each MC event, based on neutrino parent PID, px, py, pz –This does not tell us what the tweaked BMPT parameters were, but at least it allows us to determine their effect on the ND and FD CC-like event samples D. A. Petyt 11 th May 2005

2. 2/16 log 10 p z log 10 p t Robert’s slide on BMPT weighting from Argonne weight Weights are stored in 2D histograms of p_z vs p_t for each parent particle type. These weights are extracted by the BMPTHistoWeightCalculator method in MCReweight

3. 3/16 Neutrino parent – Near True energy distribution of selected CC-like events in the Near Detector by neutrino parent particle type

4. 4/16 BMPT weight vs neutrino parent Note: LOGZ scale

5. 5/16 BMPT weighting affects overall normalisation of CC-like events by – 2.8% Some structure is evident due to the different weights given to pi+, pi-, K-, p BMPT weight vs neutrino parent - 2 all pi+ pi- K+ p pbar K- K0 PIONS KAONS PROTON/PBAR

6. 6/16 X-sec and total weight True X-sec parameters were –Ma_qel=1.0123 (nominal 1.032) –Ma_res=1.0605 (nominal 1.032) –Disfact=1.0251 (nominal 1.0) Overall effect of x-sec reweighting: 1.2% change in normalisation BMPT reweighting changes normalisation by –2.8% Overall effect is –1.6%. This is why we saw no obvious excursions in any of the MDC/MC matchups with ND data. QEL RES DIS

7. 7/16 Energy dependence BMPT weight Cross-section weight BMPT * x-sec weight Visible Energy True Energy 3% change at the peak of the LE spectrum

8. 8/16 MDC/MC match-up This plot: ratio of MDC data to MC with tweaked x-sec parameters and nominal BMPT weights (i.e. 1.0). The chisq fit is to the straight line at 1.0. This plot: ratio of MDC data to MC with both tweaked x-sec and BMPT parameters. The chisq fit is also to the straight line at 1.0. The fit is acceptable, although slightly worse than before We clearly do not have enough MDC data to observe the effect of beam reweighting at this level

9. 9/16 Divining the parameters… =  ? a_pi  6% alpha_pi  5% We don’t know which BMPT parameters were changed, and by how much, but can we guess? Having gained some familiarity with the effect of BMPT weights during the MDC, I spotted the similarity between the true MDC weights returned by MCReweight (plot at left) and the effect of changing a_pi and alpha_pi in the BMPT model. Can we reproduce the left-hand plot by some combination of these two parameters?

10. 10/16 Fit to a_pi and alpha_pi Used the BMPTWeightCalculator method of MCReweight to reweight the MC events as the parameters a_pi and alpha_pi are changed. Perform a fit in the variable E vis to the ‘true’ BMPT weighted spectrum to obtain best-fit values of these parameters The fit does a reasonable job of describing the true distribution below 20 GeV. The discrepancy above that is presumably due to the effect of K- This means that a reduced parameter fit similar to that we performed for the MDC (using a_pi, alpha_pi and A_pi) would be sufficient to account for the BMPT model excursions that existed in the MDC. Unfortunately, they were so small as to be undetectable given the statistics we had. 68,90% C.L.

11. 11/16 Now perform the same analysis on FD data and see how applying the correct BMPT weights affects the oscillation fits. Neutrino parent – Far

12. 12/16 Overall shape very similar to ND sample, although proton peak is suppressed. all pi+ pi- K+ p K- K0 BMPT weight vs neutrino parent - far PIONS KAONS PROTON/PBAR

13. 13/16 Comparison of ND/FD weights ND/FD differences are rather small… FAR NEAR BMPT weight Cross-section weight BMPT * x-sec weight

14. 14/16 Far/Near ratios Effect of reweighting cancels in F/N ratio FAR NEAR Here I am plotting the product of the beam and x-sec weights…

15. 15/16 FD fit with reweighted MC The effect of re-weighting is to reduce the size of the peak in the MC sample. The oscillation fit therefore favours a slightly smaller value of sin 2 2  which moves the best-fit slightly closer to the true values. The quality of the fit does not actually improve (    versus  although both are perfectly acceptable Nominal, no osc. Nominal, best fit Weighted, no osc. Weighted, best fit NC contamination MDC ‘data’ Nominal, 68,90% C.L. Weighted, 68,90% C.L. Weighted, best fit Nominal, best fit True parameters x x o

16. 16/16 Conclusions We are now able to perform a full post-mortem of the MDC with the ability to reweight events according to true beam, x-sec and oscillation parameters The effect of beam reweighting is fairly small (overall –2.8% effect on the normalisation and ~-4% at the peak). This is partially compensated by the effect of x-sec reweighting – the overall effect is –1.6%. There is insufficient MDC data available to detect excursions that are as small as this. It appears that a reduced set of BMPT parameters is sufficient to describe the effect of beam reweighting in the MDC. Perhaps these shape and normalisation parameters are sufficiently general that they can be used outside the scope of the BMPT model? The FD fit is only marginally affected by applying the true beam and x-sec weights, the shift in parameter values is much smaller than the size of the 68% C.L contour. Not surprising given the small size of the reweighting factors applied.