update using tranquilo_ls

code/do_all.py

@@ -45,7 +45,7 @@
 still run.
 """
 
-from estimark.min import estimate_min
+from estimark.estimation import estimate
 from estimark.options import (
     all_replications,
     high_resource,
@@ -99,15 +99,15 @@ def run_replication():
     replication_specs = {}
 
     if which_model == "1" or which_model == "":
-        replication_specs["agent_name"] = "IndShock"
+        agent_name = "IndShock"
     elif which_model == "2":
-        replication_specs["agent_name"] = "Portfolio"
+        agent_name = "Portfolio"
     elif which_model == "3":
-        replication_specs["agent_name"] = "WarmGlow"
+        agent_name = "WarmGlow"
     elif which_model == "4":
-        replication_specs["agent_name"] = "WarmGlowPortfolio"
+        agent_name = "WarmGlowPortfolio"
     elif which_model == "5":
-        replication_specs["agent_name"] = "WealthPortfolio"
+        agent_name = "WealthPortfolio"
     else:
         print("Invalid model choice.")
         return
@@ -135,15 +135,23 @@ def run_replication():
         print("Invalid replication choice.")
         return
 
-    if subjective_markets == "2" or subjective_markets == "4":
-        replication_specs["subjective_stock"] = True
-        print("Adding subjective stock market beliefs...")
+    if int(subjective_markets) > 1:
+        agent_name += "Sub"
 
-    if subjective_markets == "3" or subjective_markets == "4":
-        replication_specs["subjective_labor"] = True
-        print("Adding subjective labor market beliefs...")
+        if subjective_markets == "2" or subjective_markets == "4":
+            agent_name += "(Stock)"
+            print("Adding subjective stock market beliefs...")
 
-    estimate_min(**replication_specs)
+        if subjective_markets == "3" or subjective_markets == "4":
+            agent_name += "(Labor)"
+            print("Adding subjective labor market beliefs...")
+
+        agent_name += "Market"
+
+    replication_specs["agent_name"] = agent_name
+    replication_specs["save_dir"] = "content/tables/min"
+
+    estimate(**replication_specs)
 
 
 if __name__ == "__main__":

code/estimark/estimation.py

@@ -322,6 +322,7 @@ def msm_criterion(params, agent=None, emp_moments=None):
         median wealth-to-permanent-income ratio in the simulation.
 
     """
+    emp_moments = emp_moments.copy()
     sim_moments = simulate_moments(params, agent, emp_moments)
 
     # normalize wealth moments by the maximum value in the empirical moments
@@ -521,12 +522,7 @@ def do_estimate_model(
     # Create the simple estimate table
     estimate_results_file = save_dir / (agent.name + "_estimate_results.csv")
 
-    keys_to_save = [
-        key
-        for key in vars(res)
-        if key
-        not in ["history", "convergence_report", "multistart_info", "algorithm_output"]
-    ]
+    keys_to_save = vars(res)
 
     with open(estimate_results_file, "w") as f:
         writer = csv.writer(f)
@@ -858,7 +854,7 @@ def estimate(
 if __name__ == "__main__":
     # Set booleans to determine which tasks should be done
     # Which agent type to estimate ("IndShock" or "Portfolio")
-    local_agent_name = "IndShock"
+    local_agent_name = "IndShockSub(Labor)Market"
     local_params_to_estimate = ["CRRA", "DiscFac"]
     local_estimate_model = True  # Whether to estimate the model
     # Whether to get standard errors via bootstrap

code/estimark/parameters.py

@@ -32,7 +32,7 @@
 # Number of points in discrete approximation to transitory income shocks
 TranShkCount = 7
 UnempPrb = 0.05  # Probability of unemployment while working
-UnempPrbRet = 0.005  # Probability of "unemployment" while retired # maybe one more zero
+UnempPrbRet = 0.005  # Probability of "unemployment" while retired
 IncUnemp = 0.3  # Unemployment benefits replacement rate
 IncUnempRet = 0.0  # "Unemployment" benefits when retired
 ss_variances = True  # Use the Sabelhaus-Song variance profiles
@@ -63,7 +63,7 @@
 bounds_CRRA = [1.1, 20.0]
 
 # Initial guess of the adjustment to the discount factor during estimation (beth)
-init_DiscFac = 1.0
+init_DiscFac = 0.95
 # Bounds for beth; if violated, objective function returns "penalty value"
 bounds_DiscFac = [0.5, 1.1]
 
@@ -152,17 +152,18 @@
 }
 
 minimize_options = {
-    "algorithm": "scipy_neldermead",
+    "algorithm": "tranquilo_ls",
     "multistart": True,
     "error_handling": "continue",
     "algo_options": {
         "convergence.absolute_params_tolerance": 1e-3,
         "convergence.absolute_criterion_tolerance": 1e-3,
         "stopping.max_iterations": 50,
         "stopping.max_criterion_evaluations": 100,
-        # "n_cores": 12,
+        "trustregion_initial_radius": 0.01,
+        "n_cores": 12,
     },
-    # "numdiff_options": {"n_cores": 12},
+    "numdiff_options": {"n_cores": 12},
 }
 
 # -----------------------------------------------------------------------------
@@ -219,6 +220,6 @@
 
 # from Tao's JMP
 init_subjective_labor = {
-    "TranShkStd": [0.03] * len(inc_calib["TranShkStd"]),
-    "PermShkStd": [0.03] * len(inc_calib["PermShkStd"]),
+    "TranShkStd": [0.03] * (retirement_t + 1) + [0.0] * (terminal_t - retirement_t - 1),
+    "PermShkStd": [0.03] * (retirement_t + 1) + [0.0] * (terminal_t - retirement_t - 1),
 }

code/run_all.py

@@ -1,7 +1,5 @@
 import itertools
 
-import dask
-from dask.distributed import Client
 from estimark.estimation import estimate
 from estimark.options import low_resource
 
@@ -16,10 +14,6 @@
 
 # Ask the user which replication to run, and run it:
 def run_replication():
-    client = Client(threads_per_worker=10, n_workers=20)
-
-    lazy_results = []
-
     for agent_name in agent_names:
         for sub_stock, sub_labor in itertools.product(range(2), repeat=2):
             temp_agent_name = agent_name
@@ -37,12 +31,7 @@ def run_replication():
 
             print("Model: ", replication_specs["agent_name"])
 
-            lazy_result = dask.delayed(estimate)(**replication_specs)
-            lazy_results.append(lazy_result)
-
-    dask.compute(*lazy_results)
-
-    client.close()
+            estimate(**replication_specs)
 
     print("All replications complete.")