def task_simulation_and_statistics_range_r(depends_on, produces):
"""
Task for estimating the parameters in interactive fixed effects model by using
different numbers of factors.
"""
# set parameters
simulate = json.loads(depends_on.read_text(encoding="utf-8"))
dgp_func = globals()[simulate["dgp_func"]]
all_r = simulate["all_r"]
np.random.seed(simulate["rng_seed"])
# Run the monte carlo simulation
df_sim_range_r = pd.DataFrame()
df_statistic_range_r = pd.DataFrame()
for r in all_r:
df_sim_result = simulation_coefficient(
dgp_func=dgp_func,
all_N=simulate["all_N"],
all_T=simulate["all_T"],
nsims=simulate["nsims"],
need_sde=simulate["need_sde"],
tolerance=simulate["tolerance"],
r=r,
interactive_start_value_effect=simulate[
"interactive_start_value_effect"],
within_effect=simulate["within_effect"],
**simulate["beta_true"],
)
df_statistic = statistics_coefficient(
all_N=simulate["all_N"],
all_T=simulate["all_T"],
nsims=simulate["nsims"],
df_sim_result=df_sim_result,
**simulate["beta_true"],
)
df_sim_range_r = df_sim_range_r.append(
pd.concat(
[
pd.DataFrame({"r": [r] * df_sim_result.shape[0]}),
df_sim_result
],
axis=1,
))
df_statistic_range_r = df_statistic_range_r.append(
pd.concat([
pd.DataFrame({"r": [r] * df_statistic.shape[0]}), df_statistic
],
axis=1))
# Store df_sim_range_r and df_statistic_range_r
df_sim_range_r = df_sim_range_r.reset_index(drop=True)
df_sim_range_r.to_csv(produces["sim_result"], index=False)
df_statistic_range_r = df_statistic_range_r.reset_index(drop=True)
df_statistic_range_r.to_csv(produces["statistic"], index=False)
def task_simulation_coefficient(depends_on, produces):
simulate = json.loads(depends_on.read_text(encoding="utf-8"))
np.random.seed(simulate["rng_seed"])
dgp_func = globals()[simulate["dgp_func"]]
# Run the monte carlo simulation
df_sim_result = simulation_coefficient(
dgp_func=dgp_func,
all_N=simulate["all_N"],
all_T=simulate["all_T"],
nsims=simulate["nsims"],
need_sde=simulate["need_sde"],
tolerance=simulate["tolerance"],
r=simulate["r"],
interactive_start_value_effect=simulate["interactive_start_value_effect"],
within_effect=simulate["within_effect"],
**simulate["beta_true"],
)
# Store df_sim_result with locations after each round
df_sim_result.to_csv(produces, index=False)
def test_simulation_coefficient_model3():
dgp_fun = dgp_interactive_fixed_effects_model
all_N = [10, 20]
all_T = [9, 19]
nsims = 2
df_sim_result = simulation_coefficient(
dgp_fun,
all_N,
all_T,
nsims,
need_sde=True,
interactive_start_value_effect="pooling",
within_effect="twoways",
beta1=1,
beta2=3,
mu=5,
gamma=2,
delta=4,
)
assert df_sim_result.shape == (4, 15)
assert not df_sim_result.isna().any(axis=None, skipna=False)
def test_simulation_coefficient_model4():
dgp_fun = dgp_interactive_fixed_effects_model_with_common_and_time_invariant
all_N = [10, 20]
all_T = [9, 19]
nsims = 2
df_sim_result = simulation_coefficient(
dgp_fun,
all_N,
all_T,
nsims,
need_sde=True,
interactive_start_value_effect="pooling",
within_effect="twoways",
beta1=1,
beta2=3,
mu=5,
gamma=2,
delta=4,
)
assert df_sim_result.shape == (4, 23)
na_expect = np.full(23, False)
na_expect[[11, 12, 21, 22]] = True
np.testing.assert_array_equal(
df_sim_result.isna().any(axis=0, skipna=False), na_expect)
def test_statistics_coefficient():
dgp_fun = dgp_interactive_fixed_effects_model_with_common_and_time_invariant
all_N = [10, 20]
all_T = [9, 19]
nsims = 2
df_sim_result = simulation_coefficient(
dgp_fun,
all_N,
all_T,
nsims,
need_sde=True,
interactive_start_value_effect="pooling",
within_effect="twoways",
beta1=1,
beta2=3,
mu=5,
gamma=2,
delta=4,
)
df_statistic = statistics_coefficient(all_N,
all_T,
nsims,
df_sim_result,
beta1=1,
beta2=3,
mu=5,
gamma=2,
delta=4)
assert df_statistic.shape == (2, 62)
na_expect = np.full(62, False)
na_expect[[35, 36, 40, 41, 45, 46, 50, 51, 55, 56, 60, 61]] = True
np.testing.assert_array_equal(
df_statistic.isna().any(axis=0, skipna=False), na_expect)
np.testing.assert_almost_equal(
df_statistic.loc[0, "mean_interactive.1"],
df_sim_result.loc[0:1, "beta_interactive.1"].mean(axis=None),
)
np.testing.assert_almost_equal(
df_statistic.loc[1, "mean_interactive.1"],
df_sim_result.loc[2:3, "beta_interactive.1"].mean(axis=None),
)
np.testing.assert_almost_equal(
df_statistic.loc[0, "bias_interactive.2"],
abs(df_statistic.loc[0, "mean_interactive.2"] - 3) / 3,
)
np.testing.assert_almost_equal(
df_statistic.loc[1, "sde_within.1"],
df_sim_result.loc[2:3, "sde_within.1"].mean(axis=None),
)
np.testing.assert_almost_equal(
df_statistic.loc[1, "ci_l_within.1"],
df_statistic.loc[1, "mean_within.1"] -
df_statistic.loc[1, "sde_within.1"] * 1.959963984540054,
)
np.testing.assert_almost_equal(
df_statistic.loc[1, "ci_u_within.1"],
df_statistic.loc[1, "mean_within.1"] +
df_statistic.loc[1, "sde_within.1"] * 1.959963984540054,
)
np.testing.assert_almost_equal(
df_statistic.loc[0, "rmse_interactive.2"],
np.sqrt(((df_sim_result.loc[0:1, "beta_interactive.2"] -
3)**2).mean(axis=None)),
)
