 |
 |
Hello, after scratching my head for 1 day I found why the reweighting at generator level was not doing what I expected it to do (i.e. just reweight the mX distribution). The problem is that the way Sven implemented it, first k+ was extracted and then the value of mx is extracted and reweighted according to the given weights. This means that if there is a correlation (as there is) between the distribution of k+ and mX, the final distribution of mx was not just the reweighted one [ just to see this visually, if mx and k+ where completely correlated then the weights would have been completely irrelevant because he would have kept doing an hit&miss on that given value of mx until a positive answer came out ] I have now a tested code which applies the weights on the overall distribution, i.e. the k+ distribution will be distorted. I think this is what we want, please shout if you disagree. This said I applied the weights as estimated by Alessio after the Breco selection in the generator and I made sure the scaling imposed the right resonant-nonresonant admixture. The comparison of the integrals is in http://www.slac.stanford.edu/~rfaccini/phys/vub/theo/integr_gen.eps The situation is not perfect, but more than acceptable given the different conditions under which weights were estimated. At this point I see two possibilities as far as the signal requests are concerned a1) redo the weights (changing the code, which means requesting a new release) at generator level and change the .dec files a2) covert the hybrid requests into pure resonant requests, since the weighting needs to be redone anyhow at analysis level for the systematics As far as the generic is concerned we could b1) stay as we are, we will anyhow use the other files for the signal b2) correct the hybrid (needs a collaboration wide poll, but better sooner than later) I lean towards the least kaos solution (given also the use we are doing now of the MC): a2 & b1. comments are welcome (I will then extend the poll to ISL) ciao ric