add some more options and models

input/v6.3/params/bg_diverse_9-tt-only_fixed-no-income.yaml

@@ -0,0 +1,50 @@
+#  + python -u estimate_biogeme_plan_choice.py --input plan-choices-diverse_9-tt-only.csv --no-income --ascs car 0 --ascs ride 0 --ascs pt 0 --ascs bike 0 --car-util 5.485672
+#  Modes:  ['walk', 'pt', 'car', 'bike', 'ride']
+#  Number of plans:  8259
+#  Number of choices for plan:  9
+#  Modes: ['walk', 'pt', 'car', 'bike', 'ride']
+#  Number of choices: 8259
+#  Using fixed ascs {'car': 0.0, 'ride': 0.0, 'pt': 0.0, 'bike': 0.0}
+#    Using fixed utility for car 5.485672
+#    Results for model plan-choices-diverse_9-tt-only_fixed_ascs_no_income
+#  Nbr of parameters:		4
+#  Sample size:			8259
+#  Excluded data:			0
+#  Null log likelihood:		-16870.42
+#  Final log likelihood:		-19458.35
+#  Likelihood ratio test (null):		-5175.862
+#  Rho square (null):			-0.153
+#  Rho bar square (null):			-0.154
+#  Akaike Information Criterion:	38924.7
+#  Bayesian Information Criterion:	38952.78
+scoring:
+  scoringParameters:
+    - modeParams:
+        - mode: walk
+          constant: 0
+        - mode: car
+          constant: -0.5341414592094356
+          dailyMonetaryConstant: -14.30
+          dailyUtilityConstant: 5.485672
+        - mode: pt
+          constant: 0.3971116
+        - mode: bike
+          constant: -1.3538876325
+        - mode: ride
+          constant: -1.23976957093642
+advancedScoring:
+  scoringParameters:
+    - subpopulation: person
+      modeParams:
+        - mode: car
+          deltaDailyConstant: 7.213378
+          varDailyConstant: truncatedNormal
+        - mode: bike
+          deltaConstant: 1.213346
+          varConstant: normal
+        - mode: pt
+          deltaConstant: 4.036673
+          varConstant: normal
+        - mode: ride
+          deltaConstant: 0.490086
+          varConstant: normal

input/v6.3/params/mxl_manual.yaml

@@ -0,0 +1,37 @@
+#  This is a baseline defined manually, close to the standard model
+
+# daily car utility is the same as in the base model, fixed and variable portion is split 50/50
+# mean of tn is ~ 0.8, -> 4.15 / 0.8 -> 5.18
+# Other modes are assigned a standard normal distribution
+
+scoring:
+  scoringParameters:
+    - modeParams:
+        - mode: walk
+          constant: 0
+        - mode: car
+          constant: 0
+          dailyMonetaryConstant: -14.30
+          dailyUtilityConstant: 4.15
+        - mode: pt
+          constant: 0.3971116
+        - mode: bike
+          constant: -1.3538876325
+        - mode: ride
+          constant: -1.23976957093642
+advancedScoring:
+  scoringParameters:
+    - subpopulation: person
+      modeParams:
+        - mode: car
+          deltaDailyConstant: 5.18
+          varDailyConstant: truncatedNormal
+        - mode: bike
+          deltaConstant: 1
+          varConstant: normal
+        - mode: pt
+          deltaConstant: 1
+          varConstant: normal
+        - mode: ride
+          deltaConstant: 1
+          varConstant: normal

src/main/python/choicemodels/estimate_biogeme_plan_choice.py

@@ -26,6 +26,7 @@
     parser.add_argument("--ascs", help="Predefined ASCs", nargs="+", action='append', default=[])
     parser.add_argument("--car-util", help="Fixed utility for car", type=float, default=None)
     parser.add_argument("--no-mxl", help="Disable mixed logit", action="store_true")
+    parser.add_argument("--no-income", help="Don't consider the income", action="store_true")
 
     args = parser.parse_args()
 
@@ -94,7 +95,7 @@
 
     for i in range(1, ds.k + 1):
         # Price is already negative
-        u = v[f"plan_{i}_price"] * UTIL_MONEY * (ds.global_income / v["income"]) ** EXP_INCOME
+        u = v[f"plan_{i}_price"] * UTIL_MONEY * (1 if args.no_income else (ds.global_income / v["income"]) ** EXP_INCOME)
         u -= v[f"plan_{i}_pt_n_switches"]
 
         for mode in ds.modes:
@@ -124,6 +125,8 @@
         modelName += "_util_money"
     if args.ascs:
         modelName += "_fixed_ascs"
+    if args.no_income:
+        modelName += "_no_income"
 
     biogeme.modelName = modelName
     biogeme.weight = v["weight"]

src/main/python/choicemodels/estimate_mixed_plan_choice.py

@@ -33,6 +33,7 @@ def analyse_mode_share(df, modes):
     parser.add_argument("--sample", help="Use sample of choice data", type=float, default=1)
     parser.add_argument("--seed", help="Random seed", type=int, default=0)
     parser.add_argument("--mnl", help="Use MNL instead of mixed logit", action="store_true")
+    parser.add_argument("--ascs", help="Predefined ASCs", nargs="+", action='append', default=[])
 
     args = parser.parse_args()
 
@@ -42,6 +43,10 @@ def analyse_mode_share(df, modes):
                       alt_list=[f"plan_{i}" for i in range(1, ds.k + 1)], empty_val=0,
                       varying=ds.varying, alt_is_prefix=True)
 
+    fixedvars = {x + "_usage": float(y) for x, y in args.ascs}
+    if fixedvars:
+        print("Using fixed ascs", fixedvars)
+
     analyse_mode_share(df, ds.modes)
 
     MixedLogit.check_if_gpu_available()
@@ -55,14 +60,17 @@ def analyse_mode_share(df, modes):
     # utils contains the price and opportunist costs
     addit = df["utils"] - df["pt_n_switches"]
 
+    # Fix car usage to its initial mnl value, otherwise there are convergence problems
+    fixedvars["car_used"] = None
+
     if not args.mnl:
         model = MixedLogit()
         model.fit(X=df[varnames], y=df['choice'], weights=df['weight'], varnames=varnames,
                   alts=df['alt'], ids=df['custom_id'], avail=df['valid'], random_state=args.seed,
                   addit=addit,
                   # randvars={"car_used": "tn"},
                   randvars={"car_used": "tn", "bike_usage": "n", "pt_usage": "n", "ride_usage": "n"},
-                  fixedvars={"car_used": None},
+                  fixedvars=fixedvars,
                   n_draws=args.n_draws, batch_size=args.batch_size, halton=True, skip_std_errs=True,
                   optim_method='L-BFGS-B')