def test_dfgls_auto(self):
dfgls = DFGLS(self.inflation, trend='ct', method='BIC', max_lags=3)
assert_equal(dfgls.lags, 2)
assert_equal(dfgls.max_lags, 3)
assert_almost_equal(dfgls.stat, -2.9035369, DECIMAL_4)
dfgls.max_lags = 1
assert_equal(dfgls.lags, 1)
def test_dfgls_auto(self):
dfgls = DFGLS(self.inflation, trend='ct', method='BIC', max_lags=3)
assert_equal(dfgls.lags, 2)
assert_equal(dfgls.max_lags, 3)
assert_almost_equal(dfgls.stat, -2.9035369, DECIMAL_4)
dfgls.max_lags = 1
assert_equal(dfgls.lags, 1)
def test_dfgls_auto(self):
dfgls = DFGLS(self.inflation, trend="ct", method="BIC", max_lags=3)
assert_equal(dfgls.lags, 2)
assert_equal(dfgls.max_lags, 3)
assert_almost_equal(dfgls.stat, -2.9035369, DECIMAL_4)
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.max_lags = 1
assert_equal(dfgls.lags, 1)
def DickeyFullerGlsTest(data, printResults=True, trend=None, lags=None):
options_Trend = trend if trend != None else {'c','ct'}
options_Lags = lags if lags != None else {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24}
#options_LagMethod = lagMethod if lagMethod != None else {'AIC', 'BIC', 't-stat', None}
results = dict()
for column in data.columns:
print("Dickey Fuller GLS test for column: " + column)
results_Trend = dict()
for option_Trend in options_Trend:
results_Lag = dict()
for option_Lag in options_Lags:
result = DFGLS(data[column].dropna(), trend=option_Trend, lags=option_Lag)
if printResults:
result.summary()
results_Lag[option_Lag] = result
results_Trend[option_Trend] = results_Lag
results[column] = results_Trend
return results
def test_dfgls_c(self):
dfgls = DFGLS(self.inflation, trend="c", lags=0)
assert_almost_equal(dfgls.stat, -6.017304, DECIMAL_4)
dfgls.summary()
dfgls.regression.summary()
assert dfgls.trend == "c"
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.trend = "c"
assert dfgls.trend == "c"
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.trend = "ct"
assert dfgls.trend == "ct"
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.trend = "c"
assert dfgls.trend == "c"
dfgls_hm = DFGLS(self.inflation, trend="c", lags=0, low_memory=False)
assert_almost_equal(dfgls_hm.stat, -6.017304, DECIMAL_4)
dfgls_lm = DFGLS(self.inflation, trend="c", lags=0, low_memory=True)
assert_almost_equal(dfgls_lm.stat, -6.017304, DECIMAL_4)
ml = dfgls.max_lags
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.max_lags = ml
def test_dfgls_c(self):
dfgls = DFGLS(self.inflation, trend="c", lags=0)
assert_almost_equal(dfgls.stat, -6.017304, DECIMAL_4)
dfgls.summary()
dfgls.regression.summary()
assert dfgls.trend == "c"
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.trend = "ct"
assert dfgls.trend == "ct"
with pytest.warns(FutureWarning, match="Mutating unit root"):
dfgls.trend = "c"
assert dfgls.trend == "c"
def test_dfgls_auto_low_memory(self):
y = np.cumsum(self.rng.standard_normal(200000))
dfgls = DFGLS(y, trend="c", method="BIC", low_memory=None)
assert isinstance(dfgls.stat, float)
assert dfgls._low_memory
def test_dfgls_bad_trend(self):
dfgls = DFGLS(self.inflation, trend="ct", method="BIC", max_lags=3)
with pytest.raises(ValueError):
dfgls.trend = "n"
assert dfgls != 0.0
#needs to have as input the prices or the residuals or the difference of the logs of the prices (i.e.log returns)
import halfLife_burak_mod
arr = np.array(df1['spread'].values)
halflife = halfLife_burak_mod.halflife(arr[-105:])
print ("Halflife = %f" %halflife)
#df['halflife'] = df.rolling(63).apply(lambda s: halfLife_burak_mod.halflife(s))
#df['halflife'].plot()
"""
#needs to have as input the prices or the residuals or the difference of the logs of the prices (i.e. log returns)
#arch: DFGLS: if statistic is less than critical value reject null_H=rw
from arch.unitroot import DFGLS
arr = np.array(df1['spread'].values)
dfgls = DFGLS(
arr[-30:]
) #105 is the least data needed to calculate simple regression beta; 58 is the mninimum for overall accuracy of the model
dfgls.trend = 'ct' #Constant and Linear Time Trend
#dfgls.trend = 'c' #Constant trend
print(
'If the dfgls test statistic is less than the critical value and pvalue=near zero, the series is mean reverting:'
)
print(dfgls.summary().as_text())
#needs to have as input the log returns of the prices or the difference of the residuals
#or difference of the log (p) -log (p.shift(1))
from vratio_hu import LoMac
df1["spread_DIFF"] = df1["spread"] - df1["spread"].shift(1)
df1 = df1.dropna()
arr = np.array(df1["spread_DIFF"][-300:].values)
def test_dfgls(self):
dfgls = DFGLS(self.inflation, trend='ct', lags=0)
assert_almost_equal(dfgls.stat, -6.300927, DECIMAL_4)
dfgls.summary()
dfgls.regression.summary()
if adf_p >= 0.05 and kpss_p <= 0.05:
case = 1
suggested_type_store.append(type_+1) # Try differencing
elif adf_p <= 0.05 and kpss_p >= 0.05:
case = 2
suggested_type_store.append(type_)
elif adf_p>=0.05 and kpss_p>=0.05:
case = 3
suggested_type_store.append('BAD THERE IS TREND')
elif adf_p<=0.05 and kpss_p<=0.05:
case = 4
suggested_type_store.append(type_ + 1) # Try differencing
results_dict[var]['stationary case'] = case
# DF-GLS test
dfgls = DFGLS(ts.copy())
results_dict[var]['dfgls p_value'] = dfgls.pvalue
# Yue wang modified Mann-Kendall tests account for serial autocorrelation. Null: No monotonic trend
mk_test = mk.yue_wang_modification_test(ts.copy())
results_dict[var]['ywmk p_value'] = mk_test.p
results_dict[var]['ywmk slope'] = mk_test.slope
results_dict[var]['ywmk trend'] = mk_test.trend
ret = pd.DataFrame.from_dict(results_dict, orient='index').reset_index()
ret['type'] = data_type_store
ret['suggested type'] = suggested_type_store
count_non2 = (ret['stationary case'] != 2).sum()
count_trend = (ret['ywmk trend'] != 'no trend').sum()
ret['type'] = data_type_store
ret = pd.concat((pd.DataFrame.from_dict({'fail count':{'stationary case':count_non2, 'ywmk trend': count_trend}}, orient='index'), ret), axis=0)
garch_plot1(lh['Close'])
print('Lean Hogs Future skewness is {}'.format(lh.skew(axis=0)[0]))
print('Lean Hogs Future kurtosis is {}'.format(lh.kurtosis(axis=0)[0]))
sns.distplot(lh['Close'], color='blue') #density plot
plt.title('1986–2018 Lean Hogs Future return frequency')
plt.xlabel('Possible range of data values')
# Pull up summary statistics
print(lh.describe())
adf = ADF(lh['Close'])
print(adf.summary().as_text())
kpss = KPSS(lh['Close'])
print(kpss.summary().as_text())
dfgls = DFGLS(lh['Close'])
print(dfgls.summary().as_text())
pp = PhillipsPerron(lh['Close'])
print(pp.summary().as_text())
za = ZivotAndrews(lh['Close'])
print(za.summary().as_text())
vr = VarianceRatio(lh['Close'], 12)
print(vr.summary().as_text())
from arch import arch_model
X = 100 * lh
import datetime as dt
am = arch_model(X, p=4, o=0, q=0, vol='Garch', dist='StudentsT')
res = am.fit(last_obs=dt.datetime(2003, 12, 31))
def test_dfgls_c(self):
dfgls = DFGLS(self.inflation, trend="c", lags=0)
assert_almost_equal(dfgls.stat, -6.017304, DECIMAL_4)
dfgls.summary()
dfgls.regression.summary()
figsize=(8, 7),
title=[
'CME Lean Hogs Future Close Price',
'CME Lean Hogs Future Annualized Volatility',
'CME Lean Hogs Future Return'
])
#plt.savefig(files.image_path + '\LH_close_vol_return.png')
data = lh['Close'].resample('M').mean() # resample daily data to monthly data
data = data['1992':'2004']
data = np.log(data / data.shift(1)).dropna() * 100 # d 1
adf = ADF(data)
print(adf.summary().as_text())
kpss = KPSS(data)
print(kpss.summary().as_text())
dfgls = DFGLS(data)
print(dfgls.summary().as_text())
pp = PhillipsPerron(data)
print(pp.summary().as_text())
za = ZivotAndrews(data)
print(za.summary().as_text())
vr = VarianceRatio(data, 12)
print(vr.summary().as_text())
print(data.describe())
print('Lean Hogs Future skewness is {}'.format(data.skew(axis=0)))
print('Lean Hogs Future kurtosis is {}'.format(data.kurtosis(axis=0)))
import matplotlib.gridspec as gridspec
import statsmodels.api as sm
import scipy.stats as stats
def test_dfgls(self):
dfgls = DFGLS(self.inflation, trend='ct', lags=0)
assert_almost_equal(dfgls.stat, -6.300927, DECIMAL_4)
dfgls.summary()
dfgls.regression.summary()
def test_dfgls_bad_trend(self):
dfgls = DFGLS(self.inflation, trend='ct', method='BIC', max_lags=3)
with pytest.raises(ValueError):
dfgls.trend = 'nc'
assert dfgls != 0.0
def test_dfgls_bad_trend(self):
dfgls = DFGLS(self.inflation, trend='ct', method='BIC', max_lags=3)
with pytest.raises(ValueError):
dfgls.trend = 'nc'
assert dfgls != 0.0