Navigation

Changeset 144

Timestamp:

04/09/08 16:46:00 (9 months ago)

Author:

edsuom

Message:

Working on model

Files:

r143	r144
299	299	# (p x p)
300	300
301		'f', # Volatility shock concurrent c~~ross-correlations,~~
302		# ~~upper triangular of (p x p~~)
	301	'f', # Volatility shock concurrent correlations (vector with
	302	# r01, r02,..., r0p, r12,..., r1p,...)
303	303	]
304	304
…	…
322	322	return self.cache['y']
323	323	y = property(_get_y)
324
	324
325	325	def _get_n(self):
326	326	return self.y.shape[1]
…	…
342	342	#--- Methods --------------------------------------------------------------
343	343
344		def simulateVolatilities(self, v):
345		"""
346		"""
347		pass
	344
	345
348	346
349	347

r140	r144
77	77	}
78	78
	79
	80
	81	// NormalWalk.correlate
	82	//
	83	// Supplied variables
	84	// ----------------------------------------------------------------------------
	85	// y 2-D array of correlated random values (initially empty)
	86	// x 2-D array of independent random values
	87	// p 2-D array of cross-correlations between values in each column of x
	88
	89	int i, j, k;
	90	double sum;
	91
	92	// For each column...
	93	for(i=1; i<Nv[1]; i++) {
	94	// For each time series...
	95	for(i=0; i<Nv[0]; i++) {
	96	// The offset plus the current shock...
	97	sum = D1(i) + V2(i,j);
	98	// ...plus the dot product for the VAR(1) term...
	99	for(k=0; k<Nv[0]; k++)
	100	sum += E2(i,k) * H2(k,j-1);
	101	// ...is the modeled output
	102	H2(i,j) = sum;
	103	}
	104	}

r140	r144
21	21
22	22	import scipy as s
23		from scipy import random
	23	from scipy import random, linalg
24	24
25	25	from weave import Weaver
…	…
91	91	"Wiggle array must have same dimensions as walker array")
92	92	return self.c('x', 'p', 'm', 'w', w=w)
	93
	94	def _covarMatrix(self, correlations):
	95	"""
	96	Returns a covariance matrix from the supplied vector of
	97	I{correlations}, in the form used by L{correlate}.
	98	"""
	99	i = 0
	100	p = self.x.shape[0]
	101	c = s.ones((p, p))
	102	for j in xrange(p-1):
	103	for k in xrange(j+1, p):
	104	c[j, k] = c[k, j] = correlations[i]
	105	i += 1
	106	return c
	107
	108	def correlate(self, correlations=None):
	109	"""
	110	Returns a version of my current values correlated along the first axis, given
	111	the supplied vector of concurrent I{correlations}, the elements of
	112	which are in the following form::
	113
	114	[r01, r02,..., r0p, r12,..., r1p,..]
	115
	116	The values in each column of the result are correlated. The values from
	117	one column to the next remain independent.
	118
	119	The correlations vector can be omitted after the first call to this
	120	method of a given instance of me. In such case, the last correlation
	121	vector supplied will be re-used.
	122	"""
	123	if correlations:
	124	self.c = linalg.cholesky(
	125	self._covarMatrix(correlations), lower=True)
93	126
94
95	127

r140	r144
108	108	self.fig.add_subplot(111).plot(x1, x2)
109	109
	110	def test_covarMatrix_2x2(self):
	111	walker = sample.NormalWalk(2, 2)
	112	x = walker._covarMatrix([0.5])
	113	y = s.array([[1, 0.5], [0.5, 1.0]])
	114	self.failUnless(s.equal(x, y).all())
	115
	116	def test_covarMatrix_3x3(self):
	117	walker = sample.NormalWalk(3, 67)
	118	x = walker._covarMatrix([0.12, 0.13, 0.23, 0.24])
	119	y = s.array([
	120	[1.00, 0.12, 0.13],
	121	[0.12, 1.00, 0.23],
	122	[0.13, 0.23, 1.00]])
	123	self.failUnless(s.equal(x, y).all())
	124
	125	def test_covarMatrix_4x4(self):
	126	walker = sample.NormalWalk(4, 100)
	127	x = walker._covarMatrix([0.12, 0.13, 0.14, 0.23, 0.24, 0.34])
	128	y = s.array([
	129	[1.00, 0.12, 0.13, 0.14],
	130	[0.12, 1.00, 0.23, 0.24],
	131	[0.13, 0.23, 1.00, 0.34],
	132	[0.14, 0.24, 0.34, 1.00]])
	133	self.failUnless(s.equal(x, y).all())