pval_tau_min[out_key] = res.tau_min
pval_tau_max[out_key] = res.tau_max
pval_tau_star[out_key] = res.tau_star
header = f'''\
"""
Critical values produced by phillips-ouliaris-simulation.py
{counts}
"""
from math import inf
'''
formatted_code = header + "CV_PARAMETERS = " + format_dict(cv_params)
formatted_code += "\n\nCV_TAU_MIN = " + format_dict(cv_tau_min)
formatted_code += "\n\nPVAL_LARGE_P = " + format_dict(pval_large_p)
formatted_code += "\n\nPVAL_SMALL_P = " + format_dict(pval_small_p)
formatted_code += "\n\nPVAL_TAU_MAX = " + format_dict(pval_tau_max)
formatted_code += "\n\nPVAL_TAU_STAR = " + format_dict(pval_tau_star)
formatted_code += "\n\nPVAL_TAU_MIN = " + format_dict(pval_tau_min)
targets = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38}
fm = FileMode(target_versions=targets)
formatted_code = format_file_contents(formatted_code, fast=False, mode=fm)
with open("../phillips_ouliaris.py", "w") as po:
po.write(formatted_code)
all_results.append(data["results"])
results = np.hstack(all_results)
cols = TIME_SERIES_LENGTHS.tolist() * len(data_files)
results = pd.DataFrame(results, index=PERCENTILES / 100.0, columns=cols)
cv_approx = estimate_cv_regression(results, critical_values)
adf_z_cv_approx[trend] = [cv_approx[cv] for cv in critical_values]
pvals = fit_pval_model(results[2000], small_order=4, use_log=True)
adf_z_max[trend] = pvals.tau_max
adf_z_min[trend] = pvals.tau_min
adf_z_star[trend] = pvals.tau_star
adf_z_small_p[trend] = pvals.small_p
adf_z_large_p[trend] = pvals.large_p
formatted_code = "adf_z_min = " + format_dict(adf_z_min)
formatted_code += "\n\nadf_z_star = " + format_dict(adf_z_star)
formatted_code += "\n\nadf_z_max = " + format_dict(adf_z_max)
formatted_code += "\n\n# The small p parameters are for np.log(np.abs(stat))\n"
formatted_code += "adf_z_small_p = " + format_dict(adf_z_small_p)
formatted_code += "\n\nadf_z_large_p = " + format_dict(adf_z_large_p)
formatted_code += "\n\nadf_z_cv_approx = " + format_dict(adf_z_cv_approx)
with open("../dickey_fuller.py", "r") as cvs:
lines = cvs.readlines()
retain = []
for line in lines:
if "# Z values from" in line:
break
retain.append(line)
for i in range(all_lhs.shape[1]):
lhs = all_lhs[:, i, :].ravel()
res = OLS(lhs, rhs).fit()
params = res.params.copy()
if res.pvalues[-1] > 0.05:
params[-1] = 0.00
out[cv].append(params)
values = np.array(out[cv]).tolist()
out[cv] = [[round(val, 5) for val in row] for row in values]
final_cv[trend] = dict(out)
final_cv["n"] = final_cv["nc"]
del final_cv["nc"]
formatted_str = format_dict(final_cv)
header = """
import numpy as np
eg_num_variables = np.arange(1, 13)
"""
targets = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38}
fm = FileMode(target_versions=targets)
formatted_code = header + "CV_PARAMETERS = " + formatted_str
# %%
large_p = {}
small_p = {}
tau_max = {}