def testFrankFrozen(self):
np.random.seed(123)
# Load matlab data set
stocks = np.loadtxt(dataDir + 'stocks.csv', delimiter=',')
x = stocks[:, 0]
y = stocks[:, 1]
stockModel = PairCopula(x, y, family={
'frank': 0,
})
# Try to fit all copula
stockModel.copulaTournament(verbosity=0)
# Eval the frozen model
frzU, frzV = stockModel.copulaModel.sample(40000)
# Eval a model with specified params
setU, setV = stockModel.copulaModel.sample(40000,
*stockModel.copulaParams[1])
# Ensure both frozen model and specified param model produce same result
frzModel = PairCopula(frzU, frzV)
setModel = PairCopula(setU, setV)
frzKtau, fp = frzModel.empKTau()
setKtau, sp = setModel.empKTau()
self.assertAlmostEqual(frzKtau, setKtau, delta=0.02)
self.assertAlmostEqual(fp, sp, delta=0.02)
def testBivariateBase(self):
print(
"--------------------- MULTI COPULA FIT TEST --------------------------"
)
# Load matlab data set
stocks = np.loadtxt(dataDir + 'stocks.csv', delimiter=',')
x = stocks[:, 0]
y = stocks[:, 1]
stockModel = PairCopula(x, y)
empTau = stockModel.empKTau() # kendall's tau corr coeff
empSRho = stockModel.empSRho() # spearmans corr coeff
empPRho = stockModel.empPRho() # pearson's corr coeff
print("Emprical kTau, spearmanr, pearsonr: " + str(empTau[0]) + ", " +
str(empSRho[0]) + ", " + str(empPRho[0]))
# Try to fit all copula
stockModel.copulaTournament()
# Ensure that the gaussian copula was chosen as the best fit
self.assertTrue(stockModel.copulaModel.name == "gauss")
self.assertTrue(stockModel.copulaParams[0] == "gauss")
# Check gaussian copula parameters for correctness
self.assertAlmostEqual(stockModel.copulaParams[1], 0.73874003, 4)
# Test kendalls criterion on original data
stockModel.copulaTournament(criterion='Kc', log=True)
# When using kendalls criterion the predicted copula should be gumbel
self.assertTrue(stockModel.copulaModel.name == "gumbel")
self.assertTrue(stockModel.copulaParams[0] == "gumbel")
self.assertTrue(stockModel.copulaModel.rotation == 0)
# Rotate the data and test kendalls criteron again
stockModelNegative = PairCopula(-1 * x, y)
empTau = stockModelNegative.empKTau()[0]
self.assertTrue(empTau < 0)
# Test kendalls criterion on original data
stockModelNegative.copulaTournament(criterion='Kc')
# When using kendalls criterion the predicted copula should be gumbel
self.assertTrue(stockModelNegative.copulaModel.name == "gumbel")
self.assertTrue(stockModelNegative.copulaParams[0] == "gumbel")
self.assertTrue(stockModelNegative.copulaModel.rotation == 1)
# Rotate the data and test kendalls criteron again
stockModelNegative = PairCopula(x, -1 * y)
empTau = stockModelNegative.empKTau()[0]
self.assertTrue(empTau < 0)
# Test kendalls criterion on original data
stockModelNegative.copulaTournament(criterion='Kc')
# When using kendalls criterion the predicted copula should be gumbel
self.assertTrue(stockModelNegative.copulaModel.name == "gumbel")
self.assertTrue(stockModelNegative.copulaParams[0] == "gumbel")
self.assertTrue(stockModelNegative.copulaModel.rotation == 3)
def testGaussFrozen(self):
# Load matlab data set
stocks = np.loadtxt(dataDir + 'stocks.csv', delimiter=',')
x = stocks[:, 0]
y = stocks[:, 1]
stockModel = PairCopula(x, y)
# Try to fit all copula
stockModel.copulaTournament(verbosity=0)
# Ensure that the gaussian copula was chosen as the best fit
self.assertTrue(stockModel.copulaModel.name == "gauss")
self.assertTrue(stockModel.copulaParams[0] == "gauss")
# Check gaussian copula parameters for correctness
self.assertAlmostEqual(stockModel.copulaParams[1][0], 0.73874003, 4)
# Eval the frozen model
frzU, frzV = stockModel.copulaModel.sample(40000)
# Eval a model with specified params
setU, setV = stockModel.copulaModel.sample(40000, (0.73874003, ))
# Ensure both frozen model and specified param model produce same result
frzModel = PairCopula(frzU, frzV)
setModel = PairCopula(setU, setV)
frzKtau, fp = frzModel.empKTau()
setKtau, sp = setModel.empKTau()
self.assertAlmostEqual(frzKtau, setKtau, delta=0.02)
self.assertAlmostEqual(fp, sp, delta=0.02)
# Eval a model with different specified params
setU2, setV2 = stockModel.copulaModel.sample(20000, (0.3, ))
setModel2 = PairCopula(setU2, setV2)
setKtau2, sp2 = setModel2.empKTau()
self.assertTrue(setKtau2 != setKtau)
self.assertTrue(abs(setKtau2 - setKtau) > 0.2)