Unblinding Updates

analysis/wwz/make_datacards.py

@@ -252,27 +252,28 @@ def handle_per_year_systs_for_fr(in_dict,year_name,do_jec27):
 #   - Replace the value with SMALL
 #   - And add |value| to the stat error to be more conservative
 #   - Shift the up/down variations to be centered around SMALL (does not touch stat uncertainty on up/down)
-def handle_negatives(in_dict):
+def handle_negatives(in_dict,zero_low_mc):
     out_dict = copy.deepcopy(in_dict)
     for cat in in_dict:
         for proc in in_dict[cat]["nominal"]:
             val = in_dict[cat]["nominal"][proc][0]
             var = in_dict[cat]["nominal"][proc][1]
             if val <= 0:
-                #!!!OLD METHOD FOR HANDLING LOW YIELD BINS!!!
-                #print(f"WARNING: Process \"{proc}\" in cat \"{cat}\" is negative ({val}), replacing with {SMALL} and shifting up/down systematic variations accordingly.")
-                #out_dict[cat]["nominal"][proc][0] = SMALL
-                #out_dict[cat]["nominal"][proc][1] = (abs(val) + np.sqrt(var))**2
-                #for syst in out_dict[cat]:
-                #    if syst == "nominal": continue # Already handled this one
-                #    syst_var_orig = out_dict[cat][syst][proc][0] # Dont bother messsing with mc stat error on the syst variation
-                #out_dict[cat][syst][proc][0] = (syst_var_orig - val) + SMALL # Center around SMALL
-                print(f"WARNING: Process \"{proc}\" in cat \"{cat}\" is negative ({val}), replacing with {SMALL} and setting variations to 1/1.")
-                out_dict[cat]["nominal"][proc][0] = SMALL
-                out_dict[cat]["nominal"][proc][1] = 0
-                for syst in out_dict[cat]:
-                    if syst == "nominal": continue # Already handled this one
-                    out_dict[cat][syst][proc][0] = SMALL
+                if not zero_low_mc:
+                    print(f"WARNING: Process \"{proc}\" in cat \"{cat}\" is negative ({val}), replacing with {SMALL} and shifting up/down systematic variations accordingly.")
+                    out_dict[cat]["nominal"][proc][0] = SMALL
+                    out_dict[cat]["nominal"][proc][1] = (abs(val) + np.sqrt(var))**2
+                    for syst in out_dict[cat]:
+                        if syst == "nominal": continue # Already handled this one
+                        syst_var_orig = out_dict[cat][syst][proc][0] # Dont bother messsing with mc stat error on the syst variation
+                        out_dict[cat][syst][proc][0] = (syst_var_orig - val) + SMALL # Center around SMALL
+                elif zero_low_mc:
+                    print(f"WARNING: Process \"{proc}\" in cat \"{cat}\" is negative ({val}), replacing with {SMALL} and setting variations to 1/1.")
+                    out_dict[cat]["nominal"][proc][0] = SMALL
+                    out_dict[cat]["nominal"][proc][1] = 0
+                    for syst in out_dict[cat]:
+                        if syst == "nominal": continue # Already handled this one
+                        out_dict[cat][syst][proc][0] = SMALL
 
     return out_dict
 
@@ -556,6 +557,7 @@ def main():
     parser.add_argument("--bdt",action="store_true",help="Use BDT SR bins")
     parser.add_argument("--jec-do-twentyseven",action="store_true",help="Use the 27 JEC uncertainty variations :(")
     parser.add_argument("--unblind",action="store_true",help="If set, use real data, otherwise use asimov data")
+    parser.add_argument("--zero-low-mc",action="store_true",help="If set, mc processes that are consistent with zero will be set to zero")
     parser.add_argument('-u', "--run", default='run2', help = "Which years to process", choices=["run2","run3","y22","y23"])
 
     args = parser.parse_args()
@@ -566,6 +568,7 @@ def main():
     do_jec27= args.jec_do_twentyseven
     use_bdt_sr = args.bdt
     unblind = args.unblind
+    zero_low_mc = args.zero_low_mc
     run = args.run
 
     # Check args
@@ -601,7 +604,7 @@ def main():
     # Get yield dictionary (nested in the order: year,cat,syst,proc)
     yld_dict_mc_allyears = {}
     for year in sample_names_dict_mc:
-        yld_dict_mc_allyears[year] = yt.get_yields(histo,sample_names_dict_mc[year])
+        yld_dict_mc_allyears[year] = yt.get_yields(histo,sample_names_dict_mc[year],zero_low_mc = zero_low_mc)
     if do_nuis:
         handle_per_year_systs_for_fr(yld_dict_mc_allyears,run,do_jec27)
 
@@ -636,7 +639,7 @@ def main():
 
 
     # Get rid of negative yields (and recenter syst variations around SMALL), should happen before computing kappas
-    yld_dict_mc = handle_negatives(yld_dict_mc)
+    yld_dict_mc = handle_negatives(yld_dict_mc,zero_low_mc)
 
     # Get the syst ratios to nominal (i.e. kappas)
     kappa_dict = None

ewkcoffea/modules/yield_tools.py

@@ -34,7 +34,7 @@ def valvar_op(valvar_1, valvar_2, op):
 ############## Getting yields from a histo ##############
 
 # Get the yields (nested in the order: year,cat,syst,proc)
-def get_yields(histo,sample_dict,blind=True,systematic_name=None):
+def get_yields(histo,sample_dict,blind=True,zero_low_mc = False,systematic_name=None):
 
     yld_dict = {}
 
@@ -58,7 +58,7 @@ def get_yields(histo,sample_dict,blind=True,systematic_name=None):
                     var = sum(sum(histo[{"category":cat_name,"process":sample_dict[proc_name],"systematic":syst_name }].variances(flow=True)))
 
                     # Zero out bins that are consistent with zero in the nominal
-                    if (np.sqrt(var_n) >= val_n):
+                    if (np.sqrt(var_n) >= val_n) and zero_low_mc and ("data" not in proc_name):
                         val = 0
                         var = 0