def test_first_generation(self):
"""
Test first generation intra-bar features
"""
roll_measure = get_roll_measure(self.data.close, window=20)
roll_impact = get_roll_impact(self.data.close, self.data.cum_dollar, window=20)
corwin_schultz = get_corwin_schultz_estimator(self.data.high, self.data.low, window=20)
bekker_parkinson = get_bekker_parkinson_vol(self.data.high, self.data.low, window=20)
# Shape assertions
self.assertEqual(self.data.shape[0], roll_measure.shape[0])
self.assertEqual(self.data.shape[0], roll_impact.shape[0])
self.assertEqual(self.data.shape[0], corwin_schultz.shape[0])
self.assertEqual(self.data.shape[0], bekker_parkinson.shape[0])
# Roll measure/impact assertions
self.assertAlmostEqual(roll_measure.max(), 7.1584, delta=1e-4)
self.assertAlmostEqual(roll_measure.mean(), 2.341, delta=1e-3)
self.assertAlmostEqual(roll_measure[25], 1.176, delta=1e-3) # Test some random value
self.assertAlmostEqual(roll_impact.max(), 1.022e-7, delta=1e-7)
self.assertAlmostEqual(roll_impact.mean(), 3.3445e-8, delta=1e-7)
self.assertAlmostEqual(roll_impact[25], 1.6807e-8, delta=1e-4)
# Test Corwin-Schultz
self.assertAlmostEqual(corwin_schultz.max(), 0.01652, delta=1e-4)
self.assertAlmostEqual(corwin_schultz.mean(), 0.00151602, delta=1e-4)
self.assertAlmostEqual(corwin_schultz[25], 0.00139617, delta=1e-4)
self.assertAlmostEqual(bekker_parkinson.max(), 0.018773, delta=1e-4)
self.assertAlmostEqual(bekker_parkinson.mean(), 0.001456, delta=1e-4)
self.assertAlmostEqual(bekker_parkinson[25], 0.000517, delta=1e-4)
# kurtosis
data['kurtosis_60'] = np.log(data['close']).diff().rolling(
window=60, min_periods=60, center=False).kurt()
data['kurtosis_30'] = np.log(data['close']).diff().rolling(
window=30, min_periods=30, center=False).kurt()
data['kurtosis_15'] = np.log(data['close']).diff().rolling(
window=15, min_periods=15, center=False).kurt()
data['kurtosis_10'] = np.log(data['close']).diff().rolling(
window=10, min_periods=10, center=False).kurt()
data['kurtosis_5'] = np.log(data['close']).diff().rolling(window=5,
min_periods=5,
center=False).kurt()
# microstructural features
data['roll_measure'] = micro.get_roll_measure(data['close'])
data['corwin_schultz_est'] = micro.get_corwin_schultz_estimator(
data['high'], data['low'], 100)
data['bekker_parkinson_vol'] = micro.get_bekker_parkinson_vol(
data['high'], data['low'], 100)
data['kyle_lambda'] = micro.get_bekker_parkinson_vol(data['close'],
data['volume'])
data['amihud_lambda'] = micro.get_bar_based_amihud_lambda(
data['close'], data['volume'])
data['hasbrouck_lambda'] = micro.get_bar_based_hasbrouck_lambda(
data['close'], data['volume'])
tick_diff = data['close'].diff()
data['tick_rule'] = np.where(tick_diff != 0, np.sign(tick_diff),
np.sign(tick_diff).shift(periods=-1))
### REMOVE NAN FOR INDICATORS
def roll_measure(df, window=20):
"""The Roll measure attempts to estimate the bid-ask spread (i.e. liquidity) of an instrument"""
return get_roll_measure(df["Close"], window)