def __init__(self, paras):
self.jobname = paras['jobname']
self.xVars = paras['xVars']
self.xInitialGuess = paras['xInitialParaCombo'] # np.array
self.targetVars = paras['targetVars']
self.target_fx = np.array(paras['target_fx'])
self.convCrit = paras['convCrit']
self.converged = False
self.makePlots = paras['plotting']
self.nUpdates = 0
self.optimFolder = paras['optimFolder']
self.x = self.xInitialGuess
self.fx = np.array([])
self.nIntpoints_max = 4
self.maxDist = 10 # maximal distance from second best point
self.protected_ix = False
self.optimRange = (-1.,0.)
self.plotRange = (-1.,0.)
# Set up logger
self.logger = wrapLogger(loggerName = 'costlyOptimizationLogger' + self.optimFolder, streamVerb = 'DEBUG',
logFile = os.path.join(self.optimFolder, 'costlyOpt.log'))
self.logger.debug('initialized new instance of costlyOptimization for jobname = %s' % paras['jobname'])
def __init__(self, paras):
self.jobname = paras["jobname"]
self.xVars = paras["xVars"]
self.xInitialGuess = paras["xInitialParaCombo"] # np.array
self.targetVars = paras["targetVars"]
self.target_fx = np.array(paras["target_fx"])
self.convCrit = paras["convCrit"]
self.converged = False
self.makePlots = paras["plotting"]
self.nUpdates = 0
self.optimFolder = paras["optimFolder"]
self.x = self.xInitialGuess
self.fx = np.array([])
self.nIntpoints_max = 4
self.maxDist = 10 # maximal distance from second best point
self.protected_ix = False
self.optimRange = (-1.0, 0.0)
self.plotRange = (-1.0, 0.0)
# Set up logger
self.logger = wrapLogger(
loggerName="costlyOptimizationLogger" + self.optimFolder,
streamVerb="DEBUG",
logFile=os.path.join(self.optimFolder, "costlyOpt.log"),
)
self.logger.debug("initialized new instance of costlyOptimization for jobname = %s" % paras["jobname"])
def lucasOneAgent(shockMatrix, transMatrix, beta, g, Etg, PD, PB, markovFilePath, deterministic = False):
'''
markovFilePath: path to parameters.in file
determistic: boolean indicating whether to compute special determistic or stochastic case.
One agent economy. Therefore we have:
C_t = Y_t
C_{t+1} = Y_{t+1}
In the normalied world:
c_t = 1
c_{t+1} = 1
'''
logFile = os.path.join(os.getcwd(), 'output/logs/lucasOneAgent.log')
lucasLogger = wrapLogger(loggerName = 'lucasOneAgentLog', streamVerb = verb, logFile = logFile)
gamma = float(getParameter(markovFilePath, 'gamma', 'bar-separated'))
psi = float(getParameter(markovFilePath, 'psi', 'bar-separated'))
cBar = float(getParameter(markovFilePath, 'cBar', 'bar-separated'))
mkov = MkovM(shockMatrix, transMatrix)
lucasLogger.debug('\n')
lucasLogger.info('log file written to %s\n' % logFile)
#np.set_printoptions(precision = 7, linewidth = 300)
if deterministic:
shockMatrix = np.array([[g]])
transMatrix = np.array([[1]])
states = len(shockMatrix)
c = np.ones(states)
rho = 1. / psi
gammaSeq = [ gamma ] * len(PD)
if psi < 0: # CRRA case
lucasLogger.debug('Using CRRA Euler Equations')
# infinitely lived stock (lucas tree)
qS_init = np.ones(states)
def compute_qS(qS):
euler_qS = eulerCRRA(g, beta, gamma, transMatrix, PD + qS) - qS
return euler_qS
qS = scipy.optimize.newton_krylov(compute_qS, qS_init)
# Price remaining assets
for state in range(len(shockMatrix)):
qD = eulerCRRA(g, beta, gamma, transMatrix, payoff = PD)
qB = eulerCRRA(g, beta, gamma, transMatrix, payoff = PB)
elif psi >= 0: # Epstein Zin case
lucasLogger.debug('Using Epstein Zin Euler Equations')
# guess initial value function
V_init = g
lucasLogger.debug('Initial V: %s ' % V_init)
V = V_init
lucasLogger.info('starting backward recursion: ')
dif = 1.
counter = 0
while dif > 1.e-10:
newV = EpsteinZin(c - cBar, g, V, beta, gamma, psi, transMatrix)
dif = np.sum(np.abs(newV - V))
V = newV.copy()
lucasLogger.debug('iteration: %d dif: %g V: %s' % ( counter, dif, V))
counter += 1
lucasLogger.info('converged after %d iterations: ' % counter)
lucasLogger.info('converged value function: %s' % V)
# infinitely lived stock
qS_init = np.ones(states)
def compute_qS(qS):
euler_qS = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PD + qS) - qS
return euler_qS
qS = scipy.optimize.newton_krylov(compute_qS, qS_init)
# Price remaining assets
for state in range(len(shockMatrix)):
qD = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PD)
qB = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PB)
# Generate output:
# stock
lucasLogger.info('')
lucasLogger.info('qS: %s' % qS)
EtPS = inner1d(transMatrix, ( PD + qS ) * g)
lucasLogger.info('EtPS: %s ' % EtPS)
EtRS = EtPS / qS
lucasLogger.info('EtRS: %s' % EtRS)
if not deterministic:
ERS = np.dot(mkov.getlmbda(), EtRS)
lucasLogger.info('ERS: %s' % ERS)
# bond
lucasLogger.info('')
lucasLogger.info('qB: %s' % qB)
# EtPB = np.dot(transMatrix, PB * g)
EtPB = inner1d(transMatrix, PB * g)
lucasLogger.info('EtPB: %s' % EtPB)
EtRB = EtPB / qB
lucasLogger.info('EtRB: %s' % EtRB)
if not deterministic:
ERB = np.dot(mkov.getlmbda(), EtRB)
lucasLogger.info('ERB: %s' % ERB)
# dividend asset
lucasLogger.info('')
lucasLogger.info('qD: %s' % qD)
EtPD = np.dot(transMatrix, PD * g)
lucasLogger.info('EtPD: %s ' % EtPD)
EtRD = EtPS / qD
lucasLogger.info('RtD: %s' % EtRD)
if not deterministic:
ERD = np.dot(mkov.getlmbda(), EtRD)
lucasLogger.info('ERD: %s' % ERD)
# risk premium
c_rp = EtRS - EtRB
lucasLogger.info('\ncond risk prem: %s' % c_rp)
if not deterministic:
rp = np.dot(mkov.getlmbda(), c_rp)
lucasLogger.info('unc risk premium: %s' % rp)
lucasLogger.debug('\ndone with riskpremium computation')
# Write prices to file
savePath = os.path.join(os.getcwd(), 'output')
lucasLogger.debug('writing qS/qB to path %slucasOneagent_qS|lucasOneagent_qS' % savePath)
np.savetxt(os.path.join(savePath, 'lucasOneAgent_qS.in'), qS, fmt = '%15.10f')
np.savetxt(os.path.join(savePath, 'lucasOneAgent_qB.in'), qB, fmt = '%15.10f')
if psi > 0:
np.savetxt(os.path.join(savePath, 'lucasOneAgent_V.in'), V, fmt = '%15.10f')
from pymods.markovChain.mcInterface import MkovM
from pymods.support.support import wrapLogger
from pymods.support.support import getParameter
from pymods.support.support import myArrayPrint
import numpy as np
import pandas
import scipy.optimize
from genMarkov import genMarkov
from numpy.core.umath_tests import inner1d
pprintFun = myArrayPrint(width = 12, prec = 7)
np.set_string_function(pprintFun, repr = False)
logger = wrapLogger(loggerName = 'lucasMainLog', streamVerb = 'DEBUG', logFile = None)
def lucasOneAgent(shockMatrix, transMatrix, beta, g, Etg, PD, PB, markovFilePath, deterministic = False):
'''
markovFilePath: path to parameters.in file
determistic: boolean indicating whether to compute special determistic or stochastic case.
One agent economy. Therefore we have:
C_t = Y_t
C_{t+1} = Y_{t+1}
In the normalied world:
c_t = 1
c_{t+1} = 1
'''
logFile = os.path.join(os.getcwd(), 'output/logs/lucasOneAgent.log')
# self.log.parent.addHandler(RedirectLoggingHandler())
# print self.log.handlers
# print self.log.parent.handlers
# print self.log.root.handlers
# self.log.critical('redirected gc3 log to ' + curFileName + '.log.')
# interface to the GC3Libs main functionality
self._core = self._get_core()
# call hook methods from derived classes
self.parse_args()
logger = wrapLogger(loggerName=curFileName + '.log',
streamVerb='INFO',
logFile=os.path.join(os.getcwd(), curFileName + '.log'))
gc3utilsLogger = wrapLogger(
loggerName='gc3' + curFileName + '.log',
streamVerb='INFO',
logFile=os.path.join(os.getcwd(), curFileName + '.log'),
streamFormat=
'{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}',
fileFormat=
'{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}')
logger.debug('hello')
def dispatch_record(record):
"""Passes a record on to the handlers on the stack. This is useful when
log records are created programmatically and already have all the
from gc3libs.cmdline import SessionBasedScript
from gc3libs.workflow import SequentialTaskCollection, ParallelTaskCollection
import gc3libs.utils
import gc3libs.debug
# logger
from pymods.support.support import wrapLogger
import pymods.support.support as support
from pymods.classes.tableDict import tableDict
from makePlots import momentPlots
# Set up logger
logger = wrapLogger(loggerName='idRiskParaSearchLogger',
streamVerb='DEBUG',
logFile=os.path.join(os.getcwd(), 'idRiskParaSearch.log'))
class solveParaCombination(SequentialTaskCollection):
def __init__(self, substs, solverParas, **sessionParas):
logger.debug('entering solveParaCombination.__init__ for job %s' %
sessionParas['jobname'])
self.iter = 0
self.jobname = sessionParas['jobname']
self.substs = substs
self.sessionParas = sessionParas
self.pathToExecutable = sessionParas['pathToExecutable']
# gc3 library imports
import gc3libs
from gc3libs import Application, Run, Task
from gc3libs.cmdline import SessionBasedScript, existing_file
import gc3libs.utils
import gc3libs.application.apppot
import gc3libs.debug
## custom application class
logger = wrapLogger(loggerName = 'gIdRisk.log', streamVerb = 'INFO', logFile = os.path.join(os.getcwd(), 'gParaSearch.log'))
class gIdRiskScript(SessionBasedScript, paraLoop):
"""
Read `.loop` files and execute the `forwardPremium` program accordingly.
"""
def __init__(self):
SessionBasedScript.__init__(
self,
version = '0.2',
# only '.loop' files are considered as valid input
input_filename_pattern = '*.loop',
)
paraLoop.__init__(self, 'INFO')
from gc3libs.cmdline import SessionBasedScript, existing_file
from gc3libs.workflow import SequentialTaskCollection, ParallelTaskCollection
import gc3libs.utils
import gc3libs.application.apppot
import gc3libs.debug
# logger
from pymods.support.support import fillInputDir
from pymods.support.support import wrapLogger
import pymods.support.support as support
from pymods.classes.tableDict import tableDict
# Set up logger
logger = wrapLogger(loggerName = 'idRiskParaSearchLogger', streamVerb = 'DEBUG', logFile = os.path.join(os.getcwd(), 'idRiskParaSearch.log'))
# call -x /home/benjamin/workspace/idrisk/model/bin/idRiskOut -b /home/benjamin/workspace/idrisk/model/base para.loop -C 10 -N -A '/home/benjamin/apppot0+ben.diskUpd.img'
class solveParaCombination(SequentialTaskCollection):
def __init__(self, substs, solverParas, **sessionParas):
logger.debug('entering solveParaCombination.__init__ for job %s' % sessionParas['jobname'])
self.iter = 0
self.jobname = 'solverParacombination' + sessionParas['jobname']
self.substs = substs
self.sessionParas = sessionParas
self.pathToExecutable = sessionParas['pathToExecutable']
#!/usr/bin/env python
import numpy as n, pylab as p, time
import os
from pymods.support.support import wrapLogger
logger = wrapLogger(loggerName='convexHullLogger',
streamVerb='INFO',
logFile=os.path.join(os.getcwd(), 'convexHull.log'))
def _angle_to_points(points, centre):
'''calculate angle in 2-D between points and x axis'''
resOut = []
for point in points:
delta = point - centre
res = n.arctan(delta[1] / delta[0])
if delta[0] < 0:
res += n.pi
resOut.append(res)
return n.array(resOut)
def _angle_to_point(point, centre):
'''calculate angle in 2-D between points and x axis'''
delta = point - centre
res = n.arctan(delta[1] / delta[0])
if delta[0] < 0:
res += n.pi
return res
from pymods.markovChain.mcInterface import mcInterface
from pymods.markovChain.mcInterface import MkovM
from pymods.support.support import wrapLogger
from pymods.support.support import getParameter
import numpy as np
import pandas
import scikits.statsmodels.tsa.api
import scikits.statsmodels.tsa
import scipy.linalg
import scipy.optimize
from pymods.support.support import myArrayPrint
pprintFun = myArrayPrint(width = 12, prec = 7)
#np.set_string_function(pprintFun, repr = False)
logger = wrapLogger(loggerName = 'genMarkovLog', streamVerb = 'DEBUG', logFile = os.path.join(os.getcwd(), 'output/logs/genMarkov.log'))
def genMarkov(markovFilePath, verb = 'INFO', nSimulation = int(5.e+4)):
logger.setStreamVerb(verb = verb)
logger.info('')
#os.system('cat ' + markovFilePath)
logger.debug('markovFilePath is %s' % markovFilePath)
# Read paramter file
beta = float(getParameter(markovFilePath, 'beta', 'bar-separated'))
muG = float(getParameter(markovFilePath, 'muG', 'bar-separated'))
sigmaG = float(getParameter(markovFilePath, 'sigmaG', 'bar-separated'))
p = float(getParameter(markovFilePath, 'p', 'bar-separated'))
dy = float(getParameter(markovFilePath, 'dy', 'bar-separated'))
from pymods.support.support import fillInputDir
from pymods.support.support import wrapLogger
# gc3 library imports
import gc3libs
from gc3libs import Application, Run, Task
from gc3libs.cmdline import SessionBasedScript, existing_file
import gc3libs.utils
import gc3libs.application.apppot
import gc3libs.debug
## custom application class
logger = wrapLogger(loggerName='gIdRisk.log',
streamVerb='INFO',
logFile=os.path.join(os.getcwd(), 'gParaSearch.log'))
class gIdRiskScript(SessionBasedScript, paraLoop):
"""
Read `.loop` files and execute the `forwardPremium` program accordingly.
"""
def __init__(self):
SessionBasedScript.__init__(
self,
version='0.2',
# only '.loop' files are considered as valid input
input_filename_pattern='*.loop',
)
paraLoop.__init__(self, 'INFO')
def createOverviewTable(resultDir, outFile = 'simulation.out', exportFileName = 'overviewSimu', sortCols = [], orderCols = [], verb = 'DEBUG'):
'''
create simple overview tables for idRisk project
'''
#set up logger
logger = wrapLogger(loggerName = 'createTableLog', streamVerb = verb, logFile = None)
# Extract the relevant result folders
resultFolders = []
for (folder, dirnames, filenames) in os.walk(resultDir):
# Walk through all subdirectories and look into those with para.* format
# (head, folderTail) = os.path.split(os.path.normpath(folder))
# if it's one of these output.1 folders, ignore
if re.search('output\.', folder):
continue
if not re.search('para.*', folder):
continue
if not outFile in filenames:
continue
fileName = os.path.join(folder, outFile)
if not os.path.getsize(fileName):
continue
resultFolder, pathToOutputFile = re.match('(.*para[^_]*_[^/]*)/(.*)', fileName).groups()
logger.debug('%s found' % fileName)
# Check if simulation.out ended properly.
# fileName = os.path.join(pathToFile[0], pathToFile[1], pathToFile[2])
outFileHandle = open(fileName)
lines = outFileHandle.read()
if lines.rfind('simulation ended'):
resultFolders.append((resultFolder, pathToOutputFile))
else:
logger.warning('simulation.out did not end properly. Check: %s' % (fileName))
if not resultFolders:
logger.debug('nothing to do, all folders empty')
return None
logger.debug('resultFolders: ')
logger.debug(resultFolders)
# initialize arrays
headers = ['run']
resultEntry = []
# Set up result matrix
logger.debug('\n')
logger.debug('Set up result matrix')
for ixFolder, (resultFolder, pathToOutputFile) in enumerate(resultFolders):
fileName = os.path.join(resultFolder, pathToOutputFile)
logger.debug('Reading folder: ' + resultFolder)
# (head, folderPath) = os.path.split(os.path.normpath(folderPath))
resultEntry.append([])
descr = re.match('.*para[^_]*_([^/]*).*', fileName).groups(1)
logger.debug('descr is ' + descr[0])
resultEntry[ixFolder].append(descr[0])
# open specific simu.out file
try:
outFileHandle = open(fileName)
except:
continue
lines = outFileHandle.readlines()
for ixLine, line in enumerate(lines):
line = line.rstrip()
if line.find('simulation ended') >= 0: continue
if line:
lineEles = line.rstrip().split(':')
if lineEles[0] == 'Iteration': continue
if re.search('iBar_Shock[2-9]Agent0', lineEles[0]): continue
if len(lineEles) != 2: continue
if len(lineEles[1].split()) > 1: # dont read vectors
continue
if len(lineEles[1].split('.')) > 2:
continue
(head, element) = [ ele.strip() for ele in lineEles]
logger.debug('head=' + head + ' element=' + element)
if ixFolder == 0:
newHead = reformatString(head)
logger.debug('head = ' + newHead)
headers.append(newHead)
if not element:
logger.debug('element is none')
elif is_int(element):
resultEntry[ixFolder].append(int(element))
elif is_float(element):
resultEntry[ixFolder].append(float(element))
else:
resultEntry[ixFolder].append(element)
logger.debug('headers = ')
logger.debug(headers)
logger.debug('')
# Create a tableDict from the read data.
overviewTableDict = tableDict.fromRowList(headers, resultEntry)
logger.debug(overviewTableDict)
# split run field
parameters = overviewTableDict['run'][0].split('_')
keys = []
for para in parameters:
para = para.split('=')
key = para[0]
if key == 'beta':
# handle the insane case where we loop over the time discount factor which has the same name
# as the slope coefficient from the UIP regression
key = 'beta_disc'
keys.append(key)
# Keys from the run field
keys = getUniqueList(keys)
logger.debug('keys = ' + str(keys))
# Set up a runs dictionary from the keys obtained from the run field.
runs = {}
runs = runs.fromkeys(keys)
for key in keys:
runs[key] = []
# Set up the keys for the final table.
# Combine keys from run column with the other column labels.
# Delete run field from relevant keys
#keys = keys + headers.tolist()
#keys.remove('run')
runs['keys'] = keys
# generate runs columns
for ixRun, run in enumerate(overviewTableDict['run']):
parameters = run.split('_')
for para in parameters:
para = para.split('=')
key = para[0]
if key == 'beta':
# handle the insane case where we loop over the time discount factor which has the same name
# as the slope coefficient from the UIP regression
key = 'beta_disc'
val = para[1]
## if run.count('_' + key) > 1:
## try:
## runs[key][ixRun].append(val)
## except IndexError:
## runs[key].append([])
## runs[key][ixRun].append(val)
## else:
runs[key].append(val)
splitRunField = tableDict.fromDict(runs)
logger.debug('\nsplit Run Field: ')
logger.debug(splitRunField)
# Add newly obtained run fields to full dictionary.
overviewTableDict.hzcatTableDict(splitRunField, append = False)
#splitRunField.hzcatTableDict(overviewTableDict)
logger.debug(overviewTableDict.cols)
logger.debug(overviewTableDict)
# Drop run field
overviewTableDict.drop('run')
# create some additional columns for analysis
overviewTableDict['eR_a_sc'] = overviewTableDict['eR_a']**4
overviewTableDict['eR_b_sc'] = overviewTableDict['eR_b']**4
overviewTableDict['rP_ana_sc'] = overviewTableDict['eR_a_sc'] - overviewTableDict['eR_b_sc']
overviewTableDict['std_R_a_sc'] = overviewTableDict['std_R_a(quar)'] * 2
overviewTableDict['std_R_b_sc'] = overviewTableDict['std_R_b(quar)'] * 2
# place new columns in orderCols
for ele in [ 'eR_a_sc', 'eR_b_sc', 'rP_ana_sc', 'std_R_a_sc', 'std_R_b_sc' ]:
if ele in orderCols: continue
orderCols.append(ele)
# cols = list(overviewTableDict.cols)
# print cols
# cols.insert(cols.index('eR_a') + 1, 'eR_a_sc')
# cols.insert(cols.index('eR_b') + 1, 'eR_b_sc')
# cols.insert(cols.index('rP_ana') + 1, 'rP_ana_sc')
# cols.insert(cols.index('std_R_a(quar)' + 1), 'std_R_a_sc')
# cols.insert(cols.index('std_R_b(quar)' + 1), 'std_R_b_sc')
# overviewTableDict.cols = np.array(cols)
if 'warning_____t_iteration_converged_no_more_than' in overviewTableDict.cols:
overviewTableDict.rename('warning_____t_iteration_converged_no_more_than', 't_maxConv')
## if 'gamma' in overviewTableDict.cols:
## overviewTableDict.sort(['gamma'])
if sortCols:
overviewTableDict.sort(sortCols)
if orderCols:
overviewTableDict.order(orderCols)
overviewTableDict.setWidth(26)
overviewTableDict.setPrec(10)
logger.debug(overviewTableDict)
logger.debug('possible sort/order columns are')
logger.debug(overviewTableDict.cols)
logger.debug('syntax: createTable.py . [sortcols] [ordercols]')
if exportFileName:
exportFilePath = os.path.join(resultDir, exportFileName)
logger.debug('Writing table to ' + exportFilePath)
overviewSimu = open(exportFilePath, 'w')
print >> overviewSimu, overviewTableDict
# flush the output in case this script is called within a larger project
# by default output gets flushed when buffer full or program exits.
overviewSimu.flush()
logger.debug('Done writing table')
return overviewTableDict
#!/usr/bin/env python
import numpy as n, pylab as p, time
import os
from pymods.support.support import wrapLogger
logger = wrapLogger(loggerName = 'convexHullLogger', streamVerb = 'INFO', logFile = os.path.join(os.getcwd(), 'convexHull.log'))
def _angle_to_points(points, centre):
'''calculate angle in 2-D between points and x axis'''
resOut = []
for point in points:
delta = point - centre
res = n.arctan(delta[1] / delta[0])
if delta[0] < 0:
res += n.pi
resOut.append(res)
return n.array(resOut)
def _angle_to_point(point, centre):
'''calculate angle in 2-D between points and x axis'''
delta = point - centre
res = n.arctan(delta[1] / delta[0])
if delta[0] < 0:
res += n.pi
return res
def _draw_triangle(p1, p2, p3, **kwargs):
def createOverviewTable(resultDir,
outFile='simulation.out',
exportFileName='overviewSimu',
sortCols=[],
orderCols=[],
verb='DEBUG'):
'''
create simple overview tables for idRisk project
'''
#set up logger
logger = wrapLogger(loggerName='createTableLog',
streamVerb=verb,
logFile=None)
# Extract the relevant result folders
resultFolders = []
for (folder, dirnames, filenames) in os.walk(resultDir):
# Walk through all subdirectories and look into those with para.* format
# (head, folderTail) = os.path.split(os.path.normpath(folder))
# if it's one of these output.1 folders, ignore
if re.search('output\.', folder):
continue
if not re.search('para.*', folder):
continue
if not outFile in filenames:
continue
fileName = os.path.join(folder, outFile)
if not os.path.getsize(fileName):
continue
resultFolder, pathToOutputFile = re.match('(.*para[^_]*_[^/]*)/(.*)',
fileName).groups()
logger.debug('%s found' % fileName)
# Check if simulation.out ended properly.
# fileName = os.path.join(pathToFile[0], pathToFile[1], pathToFile[2])
outFileHandle = open(fileName)
lines = outFileHandle.read()
if lines.rfind('simulation ended'):
resultFolders.append((resultFolder, pathToOutputFile))
else:
logger.warning('simulation.out did not end properly. Check: %s' %
(fileName))
if not resultFolders:
logger.debug('nothing to do, all folders empty')
return None
logger.debug('resultFolders: ')
logger.debug(resultFolders)
# initialize arrays
headers = ['run']
resultEntry = []
# Set up result matrix
logger.debug('\n')
logger.debug('Set up result matrix')
for ixFolder, (resultFolder, pathToOutputFile) in enumerate(resultFolders):
fileName = os.path.join(resultFolder, pathToOutputFile)
logger.debug('Reading folder: ' + resultFolder)
# (head, folderPath) = os.path.split(os.path.normpath(folderPath))
resultEntry.append([])
descr = re.match('.*para[^_]*_([^/]*).*', fileName).groups(1)
logger.debug('descr is ' + descr[0])
resultEntry[ixFolder].append(descr[0])
# open specific simu.out file
try:
outFileHandle = open(fileName)
except:
continue
lines = outFileHandle.readlines()
for ixLine, line in enumerate(lines):
line = line.rstrip()
if line.find('simulation ended') >= 0: continue
if line:
lineEles = line.rstrip().split(':')
if lineEles[0] == 'Iteration': continue
if re.search('iBar_Shock[2-9]Agent0', lineEles[0]): continue
if len(lineEles) != 2: continue
if len(lineEles[1].split()) > 1: # dont read vectors
continue
if len(lineEles[1].split('.')) > 2:
continue
(head, element) = [ele.strip() for ele in lineEles]
logger.debug('head=' + head + ' element=' + element)
if ixFolder == 0:
newHead = reformatString(head)
logger.debug('head = ' + newHead)
headers.append(newHead)
if not element:
logger.debug('element is none')
elif is_int(element):
resultEntry[ixFolder].append(int(element))
elif is_float(element):
resultEntry[ixFolder].append(float(element))
else:
resultEntry[ixFolder].append(element)
logger.debug('headers = ')
logger.debug(headers)
logger.debug('')
# Create a tableDict from the read data.
overviewTableDict = tableDict.fromRowList(headers, resultEntry)
logger.debug(overviewTableDict)
# split run field
parameters = overviewTableDict['run'][0].split('_')
keys = []
for para in parameters:
para = para.split('=')
key = para[0]
if key == 'beta':
# handle the insane case where we loop over the time discount factor which has the same name
# as the slope coefficient from the UIP regression
key = 'beta_disc'
keys.append(key)
# Keys from the run field
keys = getUniqueList(keys)
logger.debug('keys = ' + str(keys))
# Set up a runs dictionary from the keys obtained from the run field.
runs = {}
runs = runs.fromkeys(keys)
for key in keys:
runs[key] = []
# Set up the keys for the final table.
# Combine keys from run column with the other column labels.
# Delete run field from relevant keys
#keys = keys + headers.tolist()
#keys.remove('run')
runs['keys'] = keys
# generate runs columns
for ixRun, run in enumerate(overviewTableDict['run']):
parameters = run.split('_')
for para in parameters:
para = para.split('=')
key = para[0]
if key == 'beta':
# handle the insane case where we loop over the time discount factor which has the same name
# as the slope coefficient from the UIP regression
key = 'beta_disc'
val = para[1]
## if run.count('_' + key) > 1:
## try:
## runs[key][ixRun].append(val)
## except IndexError:
## runs[key].append([])
## runs[key][ixRun].append(val)
## else:
runs[key].append(val)
splitRunField = tableDict.fromDict(runs)
logger.debug('\nsplit Run Field: ')
logger.debug(splitRunField)
# Add newly obtained run fields to full dictionary.
overviewTableDict.hzcatTableDict(splitRunField, append=False)
#splitRunField.hzcatTableDict(overviewTableDict)
logger.debug(overviewTableDict.cols)
logger.debug(overviewTableDict)
# Drop run field
overviewTableDict.drop('run')
# create some additional columns for analysis
overviewTableDict['eR_a_sc'] = overviewTableDict['eR_a']**4
overviewTableDict['eR_b_sc'] = overviewTableDict['eR_b']**4
overviewTableDict['rP_ana_sc'] = overviewTableDict[
'eR_a_sc'] - overviewTableDict['eR_b_sc']
overviewTableDict['std_R_a_sc'] = overviewTableDict['std_R_a(quar)'] * 2
overviewTableDict['std_R_b_sc'] = overviewTableDict['std_R_b(quar)'] * 2
# place new columns in orderCols
for ele in ['eR_a_sc', 'eR_b_sc', 'rP_ana_sc', 'std_R_a_sc', 'std_R_b_sc']:
if ele in orderCols: continue
orderCols.append(ele)
# cols = list(overviewTableDict.cols)
# print cols
# cols.insert(cols.index('eR_a') + 1, 'eR_a_sc')
# cols.insert(cols.index('eR_b') + 1, 'eR_b_sc')
# cols.insert(cols.index('rP_ana') + 1, 'rP_ana_sc')
# cols.insert(cols.index('std_R_a(quar)' + 1), 'std_R_a_sc')
# cols.insert(cols.index('std_R_b(quar)' + 1), 'std_R_b_sc')
# overviewTableDict.cols = np.array(cols)
if 'warning_____t_iteration_converged_no_more_than' in overviewTableDict.cols:
overviewTableDict.rename(
'warning_____t_iteration_converged_no_more_than', 't_maxConv')
## if 'gamma' in overviewTableDict.cols:
## overviewTableDict.sort(['gamma'])
if sortCols:
overviewTableDict.sort(sortCols)
if orderCols:
overviewTableDict.order(orderCols)
overviewTableDict.setWidth(26)
overviewTableDict.setPrec(10)
logger.debug(overviewTableDict)
logger.debug('possible sort/order columns are')
logger.debug(overviewTableDict.cols)
logger.debug('syntax: createTable.py . [sortcols] [ordercols]')
if exportFileName:
exportFilePath = os.path.join(resultDir, exportFileName)
logger.debug('Writing table to ' + exportFilePath)
overviewSimu = open(exportFilePath, 'w')
print >> overviewSimu, overviewTableDict
# flush the output in case this script is called within a larger project
# by default output gets flushed when buffer full or program exits.
overviewSimu.flush()
logger.debug('Done writing table')
return overviewTableDict
self.log.parent.handlers = []
self.log.parent.addHandler(RedirectLoggingHandler())
print self.log.handlers
print self.log.parent.handlers
print self.log.root.handlers
self.log.critical("Successfully overridden gc3pie error handling. ")
# interface to the GC3Libs main functionality
self._core = self._make_core()
# call hook methods from derived classes
self.parse_args()
logger = wrapLogger(loggerName="ghousing.log", streamVerb="INFO", logFile=os.path.join(os.getcwd(), "ghousing.log"))
gc3utilsLogger = wrapLogger(
loggerName="gc3ghousing.log",
streamVerb="INFO",
logFile=os.path.join(os.getcwd(), "ghousing.log"),
streamFormat="{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}",
fileFormat="{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}",
)
def dispatch_record(record):
"""Passes a record on to the handlers on the stack. This is useful when
log records are created programmatically and already have all the
information attached and should be dispatched independent of a logger.
"""
# logbook.base._default_dispatcher.call_handlers(record)
sys.path.append(path2Pymods)
from pymods.classes.tableDict import tableDict
from pymods.support.support import wrapLogger
# check matplotlib plot properties (kwargs) here:
# http://matplotlib.sourceforge.net/api/pyplot_api.html#matplotlib.pyplot.plot
import matplotlib.cm
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
from createTable import createOverviewTable
verb = 'DEBUG'
logger = wrapLogger(loggerName='createTableLog', streamVerb=verb, logFile=None)
# list of characters for line style
# http://www.thetechrepo.com/main-articles/469
def momentPlots(baseName,
path,
xVar,
overlay,
conditions={},
xVarRange=None,
figureFile=None,
type='presentation'):
if os.path.isdir(path):
def tauchen(F=np.matrix([[0.1, 0], [0, 0.05]]),
F0=np.array([[0.9], [0.95]]),
Sigma=np.matrix([[0.01, 0.0120], [0.0120, 0.09]]),
Ns=3,
bandwidth=1,
verb='DEBUG'):
'''
function [P, y, x, MVndx, Moments0, p0] = tauchen(F, F0, Sigma, Ns, bandwidth)
convert VAR(1) for y into Markov-Chain using Tauchen's method
cool: Y can have arbitrary dimension!
Y_t = F0 + F Y_{t-1} + e_t and E e_t e_t' = Sigma
P is transition matrix
y and x are midpoints of the Markov grid. x being the orthogonalized processes (using a choleski of Sigma)
Ns is number of states *per variable* ! (Nstar = Ns^Ny)
bandwidth is multiple of standard deviations which will be covered per variable
Note: this function imposes same bandwidth and number of gridpoints per element of Y (makes algorithm more straightforward)
'''
# Get logger
logger = wrapLogger(loggerName='mcInterface', streamVerb=verb, logFile='')
# Make sure that input matrices are not arrays to allow for * notation
F = np.mat(F)
Sigma = np.mat(Sigma)
# construct the range over which each of the compents may vary
# Y_t = F0 + F Y_{t-1} + Q \varepsilon_t
Ny = np.size(F, 0)
Nstar = Ns**Ny
Q = np.linalg.cholesky(
Sigma
) # cholesky in Matlab is the transpose of numpy. -> Drop one transpose
iQ = np.linalg.inv(Q)
Iy = np.eye(Ny)
EY = np.linalg.inv(Iy - F) * F0
VarY = disclyap(F, Q * np.transpose(Q))
# X_t = Q^{-1} Y_t = F0x + Fx X_{t-1} + \varepsilon_t
Fx = iQ * F * Q
F0x = iQ * F0
EX = iQ * EY
VarX = iQ * VarY * iQ.T # = dlyap(F, Iy);
StdX = np.sqrt(VarX.diagonal()).T
# construct univariate grids for x (always midpoints!)
griddy = np.tile(np.nan, (Ns, Ny))
Ub = EX + bandwidth * StdX
Lb = EX - bandwidth * StdX
steps = (Ub - Lb) / (Ns - 1)
for x in range(Ny):
griddy[:, x] = np.linspace(Lb[x, 0], Ub[x, 0], Ns)
#np.arange(Lb[x, 0], Ub[x, 0], steps[x, 0])
# Index for Multivariate Grid
MVndx = gridMVndx(Ns, Ny)
#indexArr = np.array(list(getIndex([Ns, Ny], restr = None, direction = 'columnwise')))
#indexArr = np.reshape(indexArr, (Ny, Ns, Ny))
#midPattern = np.array(list(getIndex([Ns, Ns], restr = None, direction = 'columnwise')))
# note: midpoints also used for conditional means! (see below)
griddyRaveled = np.ravel(griddy, order='F')
x = griddyRaveled[MVndx]
y = x * Q.T
endpoints = griddy[:-1, :] + np.tile(steps.T / 2, (Ns - 1, 1))
# conditional distributions
# note usage of griddy, not XuvgridMid!
condmean = x * Fx.T + np.tile(F0x.T, (Nstar, 1))
condstd = np.ones((1, Ny))
P = np.tile(np.nan, (Nstar, Nstar))
for s in range(Nstar):
E = np.tile(condmean[s, :], (Ns - 1, 1))
V = np.tile(condstd, (Ns - 1, 1))
cdfValues = scipy.stats.norm.cdf(endpoints, E, V)
probby = np.diff(np.vstack((np.zeros(
(1, Ny)), cdfValues, np.ones((1, Ny)))),
axis=0)
probbyRaveled = np.ravel(probby, order='F')
P[s, :] = np.prod(probbyRaveled[MVndx], axis=1)
# construct unconditional distribution -- diagonalize VarX !!
if Ny > 1:
logger.debug('p0 not correctly implemented! need MVgrid')
colly = np.linalg.inv(np.linalg.cholesky(VarX).T)
uncondmean = (colly * EX).T
uncondstd = np.sqrt(np.diag(colly * VarX * colly.T)).T
E = np.tile(uncondmean, (Ns - 1, 1))
V = np.tile(uncondstd, (Ns - 1, 1))
cdfValues = scipy.stats.norm.cdf(endpoints * colly.T, E, V)
ProbUV = np.diff(np.vstack((np.zeros((1, Ny)), cdfValues, np.ones(
(1, Ny)))),
axis=0)
probUVRaveled = np.ravel(ProbUV, order='F')
p0 = np.prod(probUVRaveled[MVndx], axis=1)
Moments0 = {}
Moments0['EY'] = EY
Moments0['EX'] = EX
Moments0['VarY'] = VarY
Moments0['VarX'] = VarX
returnDict = {}
returnDict['P'] = P
returnDict['y'] = y
returnDict['x'] = x
returnDict['MVndx'] = MVndx
returnDict['Moments0'] = Moments0
returnDict['p0'] = p0
logger.debug('done tauchen')
return returnDict
# redirect_logging() # does the same thing as adding a RedirectLoggingHandler... might as well be explicit
self.log.parent.handlers = []
self.log.parent.addHandler(RedirectLoggingHandler())
print self.log.handlers
print self.log.parent.handlers
print self.log.root.handlers
self.log.critical('Successfully overridden gc3pie error handling. ')
# interface to the GC3Libs main functionality
self._core = self._make_core()
# call hook methods from derived classes
self.parse_args()
logger = wrapLogger(loggerName = 'ghousing.log', streamVerb = 'INFO', logFile = os.path.join(os.getcwd(), 'ghousing.log'))
gc3utilsLogger = wrapLogger(loggerName = 'gc3ghousing.log', streamVerb = 'INFO', logFile = os.path.join(os.getcwd(), 'ghousing.log'),
streamFormat = '{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}',
fileFormat = '{record.time:%Y-%m-%d %H:%M:%S} - {record.channel}: {record.message}')
def dispatch_record(record):
"""Passes a record on to the handlers on the stack. This is useful when
log records are created programmatically and already have all the
information attached and should be dispatched independent of a logger.
"""
# logbook.base._default_dispatcher.call_handlers(record)
gc3utilsLogger.call_handlers(record)
import gc3libs.cmdline
#gc3libs.cmdline._Script.pre_run = pre_run
import logbook
from pymods.markovChain.mcInterface import MkovM
from pymods.support.support import wrapLogger
from pymods.support.support import getParameter
from pymods.support.support import myArrayPrint
import numpy as np
import pandas
import scipy.optimize
from genMarkov import genMarkov
from numpy.core.umath_tests import inner1d
pprintFun = myArrayPrint(width = 12, prec = 7)
np.set_string_function(pprintFun, repr = False)
logger = wrapLogger(loggerName = 'lucasMainLog', streamVerb = 'DEBUG', logFile = None)
def lucasOneAgent(shockMatrix, transMatrix, beta, g, Etg, PD, PB, markovFilePath, deterministic = False):
'''
markovFilePath: path to parameters.in file
determistic: boolean indicating whether to compute special determistic or stochastic case.
One agent economy. Therefore we have:
C_t = Y_t
C_{t+1} = Y_{t+1}
In the normalied world:
c_t = 1
c_{t+1} = 1
'''
logFile = os.path.join(os.getcwd(), 'output/logs/lucasOneAgent.log')
from pymods.markovChain.mcInterface import MkovM
from pymods.support.support import wrapLogger
from pymods.support.support import getParameter
import numpy as np
import pandas
import scikits.statsmodels.tsa.api
import scikits.statsmodels.tsa
import scipy.linalg
import scipy.optimize
from pymods.support.support import myArrayPrint
pprintFun = myArrayPrint(width=12, prec=7)
#np.set_string_function(pprintFun, repr = False)
logger = wrapLogger(loggerName='genMarkovLog',
streamVerb='DEBUG',
logFile=os.path.join(os.getcwd(),
'output/logs/genMarkov.log'))
def genMarkov(markovFilePath, verb='INFO', nSimulation=int(5.e+4)):
logger.setStreamVerb(verb=verb)
logger.info('')
#os.system('cat ' + markovFilePath)
logger.debug('markovFilePath is %s' % markovFilePath)
# Read paramter file
beta = float(getParameter(markovFilePath, 'beta', 'bar-separated'))
muG = float(getParameter(markovFilePath, 'muG', 'bar-separated'))
sigmaG = float(getParameter(markovFilePath, 'sigmaG', 'bar-separated'))
def lucasOneAgent(shockMatrix, transMatrix, beta, g, Etg, PD, PB, markovFilePath, deterministic = False):
'''
markovFilePath: path to parameters.in file
determistic: boolean indicating whether to compute special determistic or stochastic case.
One agent economy. Therefore we have:
C_t = Y_t
C_{t+1} = Y_{t+1}
In the normalied world:
c_t = 1
c_{t+1} = 1
'''
logFile = os.path.join(os.getcwd(), 'output/logs/lucasOneAgent.log')
lucasLogger = wrapLogger(loggerName = 'lucasOneAgentLog', streamVerb = verb, logFile = logFile)
gamma = float(getParameter(markovFilePath, 'gamma', 'bar-separated'))
psi = float(getParameter(markovFilePath, 'psi', 'bar-separated'))
cBar = float(getParameter(markovFilePath, 'cBar', 'bar-separated'))
mkov = MkovM(shockMatrix, transMatrix)
lucasLogger.debug('\n')
lucasLogger.info('log file written to %s\n' % logFile)
#np.set_printoptions(precision = 7, linewidth = 300)
if deterministic:
shockMatrix = np.array([[g]])
transMatrix = np.array([[1]])
states = len(shockMatrix)
c = np.ones(states)
rho = 1. / psi
gammaSeq = [ gamma ] * len(PD)
if psi < 0: # CRRA case
lucasLogger.debug('Using CRRA Euler Equations')
# infinitely lived stock (lucas tree)
qS_init = np.ones(states)
def compute_qS(qS):
euler_qS = eulerCRRA(g, beta, gamma, transMatrix, PD + qS) - qS
return euler_qS
qS = scipy.optimize.newton_krylov(compute_qS, qS_init)
# Price remaining assets
for state in range(len(shockMatrix)):
qD = eulerCRRA(g, beta, gamma, transMatrix, payoff = PD)
qB = eulerCRRA(g, beta, gamma, transMatrix, payoff = PB)
elif psi >= 0: # Epstein Zin case
lucasLogger.debug('Using Epstein Zin Euler Equations')
# guess initial value function
V_init = g
lucasLogger.debug('Initial V: %s ' % V_init)
V = V_init
lucasLogger.info('starting backward recursion: ')
dif = 1.
counter = 0
while dif > 1.e-10:
newV = EpsteinZin(c - cBar, g, V, beta, gamma, psi, transMatrix)
dif = np.sum(np.abs(newV - V))
V = newV.copy()
lucasLogger.debug('iteration: %d dif: %g V: %s' % ( counter, dif, V))
counter += 1
lucasLogger.info('converged after %d iterations: ' % counter)
lucasLogger.info('converged value function: %s' % V)
# infinitely lived stock
qS_init = np.ones(states)
def compute_qS(qS):
euler_qS = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PD + qS) - qS
return euler_qS
qS = scipy.optimize.newton_krylov(compute_qS, qS_init)
# Price remaining assets
for state in range(len(shockMatrix)):
qD = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PD)
qB = eulerEpsteinZin(g, V, beta, gamma, psi, transMatrix, payoff = PB)
# Generate output:
# stock
lucasLogger.info('')
lucasLogger.info('qS: %s' % qS)
EtPS = inner1d(transMatrix, ( PD + qS ) * g)
lucasLogger.info('EtPS: %s ' % EtPS)
EtRS = EtPS / qS
lucasLogger.info('EtRS: %s' % EtRS)
if not deterministic:
ERS = np.dot(mkov.getlmbda(), EtRS)
lucasLogger.info('ERS: %s' % ERS)
# bond
lucasLogger.info('')
lucasLogger.info('qB: %s' % qB)
# EtPB = np.dot(transMatrix, PB * g)
EtPB = inner1d(transMatrix, PB * g)
lucasLogger.info('EtPB: %s' % EtPB)
EtRB = EtPB / qB
lucasLogger.info('EtRB: %s' % EtRB)
if not deterministic:
ERB = np.dot(mkov.getlmbda(), EtRB)
lucasLogger.info('ERB: %s' % ERB)
# dividend asset
lucasLogger.info('')
lucasLogger.info('qD: %s' % qD)
EtPD = np.dot(transMatrix, PD * g)
lucasLogger.info('EtPD: %s ' % EtPD)
EtRD = EtPS / qD
lucasLogger.info('RtD: %s' % EtRD)
if not deterministic:
ERD = np.dot(mkov.getlmbda(), EtRD)
lucasLogger.info('ERD: %s' % ERD)
# risk premium
c_rp = EtRS - EtRB
lucasLogger.info('\ncond risk prem: %s' % c_rp)
if not deterministic:
rp = np.dot(mkov.getlmbda(), c_rp)
lucasLogger.info('unc risk premium: %s' % rp)
lucasLogger.debug('\ndone with riskpremium computation')
# Write prices to file
savePath = os.path.join(os.getcwd(), 'output')
lucasLogger.debug('writing qS/qB to path %slucasOneagent_qS|lucasOneagent_qS' % savePath)
np.savetxt(os.path.join(savePath, 'lucasOneAgent_qS.in'), qS, fmt = '%15.10f')
np.savetxt(os.path.join(savePath, 'lucasOneAgent_qB.in'), qB, fmt = '%15.10f')
if psi > 0:
np.savetxt(os.path.join(savePath, 'lucasOneAgent_V.in'), V, fmt = '%15.10f')
sys.path.append(path2Pymods)
from pymods.classes.tableDict import tableDict
from pymods.support.support import wrapLogger
# check matplotlib plot properties (kwargs) here:
# http://matplotlib.sourceforge.net/api/pyplot_api.html#matplotlib.pyplot.plot
import matplotlib.cm
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
from createTable import createOverviewTable
verb = 'DEBUG'
logger = wrapLogger(loggerName = 'createTableLog', streamVerb = verb, logFile = None)
# list of characters for line style
# http://www.thetechrepo.com/main-articles/469
def momentPlots(baseName, path, xVar, overlay, conditions = {}, xVarRange = None, figureFile = None, type = 'presentation'):
if os.path.isdir(path):
# if not tableFile:
tableFile = os.path.join(path, 'overviewSimu')
elif os.path.isfile(path):
tableFile = path
else:
logger.critical('path is not a directory or a table file.')
def tauchen(F = np.matrix([[0.1, 0], [0, 0.05]]), F0 = np.array([[0.9], [0.95]]), Sigma = np.matrix([[0.01, 0.0120], [0.0120, 0.09]]),
Ns = 3, bandwidth = 1, verb = 'DEBUG'):
'''
function [P, y, x, MVndx, Moments0, p0] = tauchen(F, F0, Sigma, Ns, bandwidth)
convert VAR(1) for y into Markov-Chain using Tauchen's method
cool: Y can have arbitrary dimension!
Y_t = F0 + F Y_{t-1} + e_t and E e_t e_t' = Sigma
P is transition matrix
y and x are midpoints of the Markov grid. x being the orthogonalized processes (using a choleski of Sigma)
Ns is number of states *per variable* ! (Nstar = Ns^Ny)
bandwidth is multiple of standard deviations which will be covered per variable
Note: this function imposes same bandwidth and number of gridpoints per element of Y (makes algorithm more straightforward)
'''
# Get logger
logger = wrapLogger(loggerName = 'mcInterface', streamVerb = verb, logFile = '')
# Make sure that input matrices are not arrays to allow for * notation
F = np.mat(F)
Sigma = np.mat(Sigma)
# construct the range over which each of the compents may vary
# Y_t = F0 + F Y_{t-1} + Q \varepsilon_t
Ny = np.size(F, 0)
Nstar = Ns**Ny
Q = np.linalg.cholesky(Sigma) # cholesky in Matlab is the transpose of numpy. -> Drop one transpose
iQ = np.linalg.inv(Q)
Iy = np.eye(Ny)
EY = np.linalg.inv(Iy - F) * F0
VarY = disclyap(F, Q * np.transpose(Q));
# X_t = Q^{-1} Y_t = F0x + Fx X_{t-1} + \varepsilon_t
Fx = iQ * F * Q
F0x = iQ * F0
EX = iQ * EY
VarX = iQ * VarY * iQ.T # = dlyap(F, Iy);
StdX = np.sqrt(VarX.diagonal()).T
# construct univariate grids for x (always midpoints!)
griddy = np.tile(np.nan, (Ns, Ny))
Ub = EX + bandwidth * StdX
Lb = EX - bandwidth * StdX
steps = (Ub - Lb) / (Ns - 1)
for x in range(Ny):
griddy[:, x] = np.linspace(Lb[x,0], Ub[x,0], Ns)
#np.arange(Lb[x, 0], Ub[x, 0], steps[x, 0])
# Index for Multivariate Grid
MVndx = gridMVndx(Ns, Ny);
#indexArr = np.array(list(getIndex([Ns, Ny], restr = None, direction = 'columnwise')))
#indexArr = np.reshape(indexArr, (Ny, Ns, Ny))
#midPattern = np.array(list(getIndex([Ns, Ns], restr = None, direction = 'columnwise')))
# note: midpoints also used for conditional means! (see below)
griddyRaveled = np.ravel(griddy, order = 'F')
x = griddyRaveled[MVndx]
y = x * Q.T
endpoints = griddy[:-1,:] + np.tile(steps.T / 2, ( Ns - 1, 1) )
# conditional distributions
# note usage of griddy, not XuvgridMid!
condmean = x * Fx.T + np.tile(F0x.T, (Nstar,1) )
condstd = np.ones((1, Ny))
P = np.tile(np.nan, (Nstar, Nstar))
for s in range(Nstar):
E = np.tile(condmean[s,:], (Ns-1, 1))
V = np.tile(condstd, (Ns-1, 1))
cdfValues = scipy.stats.norm.cdf(endpoints, E, V)
probby = np.diff(np.vstack((np.zeros((1, Ny)), cdfValues, np.ones((1,Ny)))), axis = 0)
probbyRaveled = np.ravel(probby, order = 'F')
P[s,:] = np.prod(probbyRaveled[MVndx], axis = 1)
# construct unconditional distribution -- diagonalize VarX !!
if Ny > 1:
logger.debug('p0 not correctly implemented! need MVgrid')
colly = np.linalg.inv(np.linalg.cholesky(VarX).T)
uncondmean = (colly * EX).T
uncondstd = np.sqrt(np.diag(colly * VarX * colly.T)).T
E = np.tile(uncondmean, (Ns-1, 1))
V = np.tile(uncondstd, (Ns-1, 1))
cdfValues = scipy.stats.norm.cdf(endpoints * colly.T, E, V)
ProbUV = np.diff(np.vstack((np.zeros((1, Ny)), cdfValues, np.ones((1,Ny)))), axis = 0)
probUVRaveled = np.ravel(ProbUV, order = 'F')
p0 = np.prod(probUVRaveled[MVndx], axis = 1)
Moments0 = {}
Moments0['EY'] = EY
Moments0['EX'] = EX
Moments0['VarY'] = VarY
Moments0['VarX'] = VarX
returnDict = {}
returnDict['P'] = P
returnDict['y'] = y
returnDict['x'] = x
returnDict['MVndx'] = MVndx
returnDict['Moments0'] = Moments0
returnDict['p0'] = p0
logger.debug('done tauchen')
return returnDict
#!/usr/bin/env python
import os
import numpy as np
import scipy.optimize
import scipy.interpolate as si
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
import pylab as p
from pymods.support.support import wrapLogger
import convexHull
logger = wrapLogger(loggerName='iwdInterpolationLogger',
streamVerb='DEBUG',
logFile=os.path.join(os.getcwd(), 'iwdInterpolation.log'))
np.seterr(all='raise')
class iwdInterpolation(object):
'''
Inverse weighted density interpolation
'''
def __init__(self, xMat, fx, normExp=4, dx=1e-3, makePlot=False):
logger.debug('initializing new instance of iwdInterpolation')
self.center = xMat.mean(0)
angles = convexHull._angle_to_points(xMat, self.center)
sortIndices = angles.argsort()
self.angles = angles[sortIndices]
#!/usr/bin/env python
import os
import numpy as np
import scipy.optimize
import scipy.interpolate as si
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab
import pylab as p
from pymods.support.support import wrapLogger
import convexHull
logger = wrapLogger(loggerName = 'iwdInterpolationLogger', streamVerb = 'DEBUG', logFile = os.path.join(os.getcwd(), 'iwdInterpolation.log'))
np.seterr(all='raise')
class iwdInterpolation(object):
'''
Inverse weighted density interpolation
'''
def __init__(self, xMat, fx, normExp = 4, dx = 1e-3, makePlot=False):
logger.debug('initializing new instance of iwdInterpolation')
self.center = xMat.mean(0)
angles = convexHull._angle_to_points(xMat, self.center)
sortIndices = angles.argsort()
self.angles = angles[sortIndices]
self.xMat = xMat[sortIndices, :]
self.fx = fx[sortIndices]
