def __init__(self,path,filename):
new = Image.open(path+'/'+filename)
self.size = new.size
self.data = np.reshape(new.getdata(), self.size)
self.pmax = prctile(new.getdata(),p=95)
self.pmin = prctile(new.getdata(),p=5)
self.title= 'New range = %i-%i'%(self.pmin,self.pmax)
self.label= 'UT: %s\nExp. Times: %.1f sec\nTemp: %.1fC' % (new.info['UniversalTime'].strftime('%m-%d-%y %H:%M:%S'),new.info['ExposureTime'],new.info['CCDTemperature'])
def bootstrapMedian(data, N=5000):
# determine 95% confidence intervals of the median
M = len(data)
percentile = [2.5,97.5]
estimate = np.zeros(N)
for n in range(N):
bsIndex = np.random.random_integers(0,M-1,M)
bsData = data[bsIndex]
estimate[n] = mlab.prctile(bsData, 50)
CI = mlab.prctile(estimate, percentile)
return CI
def bootstrapMedian(data, N=5000):
# determine 95% confidence intervals of the median
M = len(data)
percentile = [2.5, 97.5]
estimate = np.zeros(N)
for n in range(N):
bsIndex = np.random.random_integers(0, M - 1, M)
bsData = data[bsIndex]
estimate[n] = mlab.prctile(bsData, 50)
CI = mlab.prctile(estimate, percentile)
return CI
def execute(self, seed=0):
"""Test the difference in means with bootstrapping.
Data is drawn randomly from group1 and group2, with resampling.
From these bootstraps, estimates with confidence intervals are
calculated for the mean of each group and the difference in means.
The estimated difference is positive if group2 > group1.
Sets: mean1, CI_1, mean2, CI_2, diff_estimate, diff_CI, p1, p2
p1 is the p-value estimated from the distribution of differences
p2 is the p-value from a 1-sample ttest on that distribution
"""
if len(self.data1) < self.min_bucket or len(self.data2) < self.min_bucket:
#~ raise BootstrapError(
#~ 'insufficient data in bucket in bootstrap_two_groups')
raise ValueError(
'insufficient data in bucket in bootstrap_two_groups')
if seed is not None:
np.random.seed(seed)
# Generate random samples, shape (n_boots, len(group))
self.idxs1 = np.random.randint(0, len(self.data1),
(self.n_boots, len(self.data1)))
self.idxs2 = np.random.randint(0, len(self.data2),
(self.n_boots, len(self.data2)))
# Draw from the data
self.draws1 = self.data1[self.idxs1]
self.draws2 = self.data2[self.idxs2]
# Bootstrapped means of each group
self.means1 = self.draws1.mean(axis=1)
self.means2 = self.draws2.mean(axis=1)
# CIs on group means
self.CI_1 = mlab.prctile(self.means1, (2.5, 97.5))
self.CI_2 = mlab.prctile(self.means2, (2.5, 97.5))
# Bootstrapped difference between the groups
self.diffs = self.means2 - self.means1
self.CI_diff = mlab.prctile(self.diffs, (2.5, 97.5))
# p-value
self.p_from_dist = pvalue_of_distribution(self.diffs, 0)
# save memory
del self.idxs1
del self.idxs2
del self.draws1
del self.draws2
def test_prctile():
x=[1,2,3]
assert mlab.prctile(x,50)==np.median(x)
x=[1,2,3,4]
assert mlab.prctile(x,50)==np.median(x)
ob1=[1,1,2,2,1,2,4,3,2,2,2,3,4,5,6,7,8,9,7,6,4,5,5]
p = [0, 75, 100]
expected = [1, 5.5, 9]
actual = mlab.prctile(ob1,p)
assert np.allclose( expected, actual )
for pi, expectedi in zip(p,expected):
actuali = mlab.prctile(ob1,pi)
assert np.allclose( expectedi, actuali )
def createAsaInfo(self):
'Return True on error.'
if self.showMessages:
nTdebug( "Fetching WHATIF per-atom surface accessibility info..." )
fileNames = glob.glob(os.path.join(self.whatIfDataDir, "wsvacc*.log"))
self.allWhatIfInfo = {'chains': {}}
for fileName in fileNames:
if self.readWhatIfAsaInfoFile(fileName): # fills self.allWhatIfInfo
nTerror("Failed %s when reading file." % (getCallerName()))
return True
# end for
#
# Now determine the median ASA for each
#
# whatIfInfo is used in super class whereas allWhatIfInfo was filled before.
self.whatIfInfo = self.allWhatIfInfo
d = self.whatIfInfo['chains']
# medianIndex = None
for chainCode in d.keys():
for seqKey in d[chainCode].keys():
for atomName in d[chainCode][seqKey]['atoms'].keys():
asaList = d[chainCode][seqKey]['atoms'][atomName]
asaList.sort()
# if not medianIndex:
# medianIndex = int((len(asaList) / 2.0) + 0.5) # fails with round off on single element lists.
ml = mlab.prctile(asaList,[50])
# if medianIndex < 0 or medianIndex >= len(asaList):
# nTerror("Found improper median index %s for %s" % (medianIndex, str(asaList)))
# return True
# d[chainCode][seqKey]['atoms'][atomName] = [asaList[medianIndex]] # Resetting list to only include median
d[chainCode][seqKey]['atoms'][atomName] = [ml[0]]
def comp_histo(a, **kwargs):
"""Return plot-ready histogram (h,l), with Freedman-Diaconis' choice for
optimal bin width if not fixed.
See http://en.wikipedia.org/wiki/Histogram"""
if 'bins' in kwargs:
try:
nbins = len(kwargs['bins'])
except:
nbins = kwargs['bins']
print "Using default numpy histogram: nbins=%d" % nbins
h,l = N.histogram(a, **kwargs)
else: # Define optimal binning
if 'range' in kwargs:
vmin,vmax = kwargs['range']
else:
vmin,vmax = a.min(),a.max()
# Freedman-Diaconis' choice for optimal bin width
q1,q3 = prctile(a, p=(25.,75.))
h = 2 * (q3-q1) / len(a)**(1./3.)
nbins = round( (vmax-vmin)/h )
print "Freedman-Diaconis optimal bin width: nbins=%d" % nbins
h,l = N.histogram(a, bins=nbins, **kwargs)
h = N.concatenate((h,[h[-1]])) # Complete h
#l = N.concatenate((l,[l[-1]+l[1]-l[0]])) # Not needed w/ new=True
return h,l
def harmonize_clim_in_subplots(fig=None, axa=None, clim=(None, None),
center_clim=False, trim=1):
"""Set clim to be the same in all subplots in figur
f : Figure to grab all axes from, or None
axa : the list of subplots (if f is None)
clim : tuple of desired c-limits. If either or both values are
unspecified, they are derived from the data.
center_clim : if True, the mean of the new clim is always zero
May overrule specified `clim`
trim : does nothing if 1 or None
otherwise, sets the clim to truncate extreme values
for example, if .99, uses the 1% and 99% values of the data
"""
# Which axes to operate on
if axa is None:
axa = fig.get_axes()
axa = np.asarray(axa)
# Two ways of getting new clim
if trim is None or trim == 1:
# Get all the clim
all_clim = []
for ax in axa.flatten():
for im in ax.get_images():
all_clim.append(np.asarray(im.get_clim()))
# Find covering clim and optionally center
all_clim_a = np.array(all_clim)
new_clim = [np.min(all_clim_a[:, 0]), np.max(all_clim_a[:, 1])]
else:
# Trim to specified prctile of the image data
data_l = []
for ax in axa.flatten():
for im in ax.get_images():
data_l.append(np.asarray(im.get_array()).flatten())
data_a = np.concatenate(data_l)
# New clim
new_clim = list(mlab.prctile(data_a, (100.*(1-trim), 100.*trim)))
# Take into account specified clim
try:
if clim[0] is not None:
new_clim[0] = clim[0]
if clim[1] is not None:
new_clim[1] = clim[1]
except IndexError:
print "warning: problem with provided clim"
# Optionally center
if center_clim:
new_clim = np.max(np.abs(new_clim)) * np.array([-1, 1])
# Set to new value
for ax in axa.flatten():
for im in ax.get_images():
im.set_clim(new_clim)
return new_clim
def HoeffdingRuleMarkov(beta, G, H, U, FlowNum):
"""
Estimate the K-L divergence and the threshold by use of weak convergence
----------------
beta: the false alarm rate
G: the gradient
H: the Hessian
U: a sample path of the Gaussian empirical measure
FlowNum: the number of flows
----------------
"""
_, SampNum, _ = U.shape
# Estimate K-L divergence using 2nd-order Taylor expansion
KL = []
for j in range(0, SampNum):
t = (1.0 / sqrt(FlowNum)) * np.dot(G, U[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(U[0, j, :], H), U[0, j, :])
# print t.tolist()
# break
KL.append(np.array(t.real)[0])
eta = prctile(KL, 100 * (1 - beta))
# print(KL)
# assert(1 == 2)
return eta
def createAsaInfo(self):
'Return True on error.'
if self.showMessages:
nTdebug("Fetching WHATIF per-atom surface accessibility info...")
fileNames = glob.glob(os.path.join(self.whatIfDataDir, "wsvacc*.log"))
self.allWhatIfInfo = {'chains': {}}
for fileName in fileNames:
if self.readWhatIfAsaInfoFile(
fileName): # fills self.allWhatIfInfo
nTerror("Failed %s when reading file." % (getCallerName()))
return True
# end for
#
# Now determine the median ASA for each
#
# whatIfInfo is used in super class whereas allWhatIfInfo was filled before.
self.whatIfInfo = self.allWhatIfInfo
d = self.whatIfInfo['chains']
# medianIndex = None
for chainCode in d.keys():
for seqKey in d[chainCode].keys():
for atomName in d[chainCode][seqKey]['atoms'].keys():
asaList = d[chainCode][seqKey]['atoms'][atomName]
asaList.sort()
# if not medianIndex:
# medianIndex = int((len(asaList) / 2.0) + 0.5) # fails with round off on single element lists.
ml = mlab.prctile(asaList, [50])
# if medianIndex < 0 or medianIndex >= len(asaList):
# nTerror("Found improper median index %s for %s" % (medianIndex, str(asaList)))
# return True
# d[chainCode][seqKey]['atoms'][atomName] = [asaList[medianIndex]] # Resetting list to only include median
d[chainCode][seqKey]['atoms'][atomName] = [ml[0]]
def bootstrapped_intercluster_mahalanobis(cluster1, cluster2, n_boots=1000,
fix_covariances=True):
"""Bootstrap the intercluster distance.
Returns:
m - The mean distance
CI - 95% confidence interval on the distance
distances - an array of the distances measured on each boot
"""
d_l = []
# Determine the covariance matrices, or recalculate each time
if fix_covariances:
icov1 = np.linalg.inv(np.cov(cluster1, rowvar=0))
icov2 = np.linalg.inv(np.cov(cluster2, rowvar=0))
else:
icov1, icov2 = None, None
# Bootstrap
for n_boot in range(n_boots):
# Draw
idxs1 = np.random.randint(0, len(cluster1), len(cluster1))
idxs2 = np.random.randint(0, len(cluster2), len(cluster2))
# Calculate and store
d_l.append(intercluster_mahalanobis(
cluster1[idxs1], cluster2[idxs2], icov1, icov2))
# Statistics
d_a = np.asarray(d_l)
m = np.mean(d_a)
CI = mlab.prctile(d_a, (2.5, 97.5))
return m, CI, d_a
def print_stats(name, x=None):
"Prints simple stats"
if type(name) is not StringType:
x = name
name = 'mean,stdv,rms,min,25%,median,75%,max: '
if name == '__header__':
print ''
n = (80 - len(x)) / 2
print n * ' ' + x
print n * ' ' + len(x) * '-'
print ''
print ' Name mean stdv rms min 25% median 75% max'
print ' --------- ------- ------- ------- ------- ------- ------- ------- -------'
elif name == '__sep__':
print ' --------- ------- ------- ------- ------- ------- ------- ------- -------'
elif name == '__footer__':
print ' --------- ------- ------- ------- ------- ------- ------- ------- -------'
print ''
else:
ave = x.mean()
std = x.std()
rms = sqrt(ave * ave + std * std)
prc = prctile(x)
print '%10s %7.2f %7.2f %7.2f %7.2f %7.2f %7.2f %7.2f %7.2f '%\
(name,ave,std,rms,prc[0],prc[1],prc[2],prc[3],prc[4])
def HoeffdingRuleMarkovRobust_(beta, G_list, H_list, U_list, FlowNum):
"""
Estimate the K-L divergence and the threshold by use of weak convergence
----------------
beta: the false alarm rate
G: the gradient
H: the Hessian
U: a sample path of the Gaussian empirical measure
FlowNum: the number of flows
----------------
"""
_, SampNum, _ = U_list[0].shape
# print(SampNum)
# assert(1 == 2)
# Estimate K-L divergence using 2nd-order Taylor expansion
KL = []
for j in range(0, SampNum):
KL_est_list = []
for G, H, U in zip(G_list, H_list, U_list):
KL_est = (1.0 / sqrt(FlowNum)) * np.dot(G, U[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(U[0, j, :], H), U[0, j, :])
KL_est = np.array(KL_est.real)[0]
KL_est_list.append(KL_est)
KL.append(min(KL_est_list))
eta = prctile(KL, 100 * (1 - beta))
return eta
def compute(self):
"""Detect RFI
"""
median_size = (self.median_size_time, self.median_size_freq)
data = self.data - sp_dip.median_filter(self.data, size=median_size)
# data1 = np.abs(np.sum(data,0))
# data1 = np.abs(np.median(data,0))
data1 = (np_median(data, 0))
# th = np.percentile(data1, th_prctile)
thl = []
for ii in xrange(data1.shape[0] - 9):
thl.append(prctile(data1[ii:ii + 10], p=90))
# thl.append(max(data1[ii:ii+10]))
th = self.th_k * np_median(thl)
for ii in xrange(data1.shape[0]):
if data1[ii] > th:
z, p_value = sp_normaltest(data[:, ii])
if p_value < self.p_th:
if self.is_out_selected('Not_normal'):
self.flag_results['Not_normal'].flag_data[:, ii] = 1
else:
if self.is_out_selected('Normal'):
self.flag_results['Normal'].flag_data[:, ii] = 1
return self.flag_results
def calc_kde1d(Xin, N=256, range=None, Verbose=False, name=None):
"""
Calculates 1D KDE. On input,
Xin --- input data array
N --- number of bins to evaluate KDE
range --- range of Xin values to work with
Example:
bins, P = calc_kde1d(obs,range=(-2,2))
"""
try:
X = Xin.data[Xin.mask == False].ravel()
except:
X = Xin
if range == None:
prc = prctile(X)
range = [prc[0], prc[4]]
bins = linspace(range[0], range[1], N)
if Verbose:
if name != None:
print name
print 'Evaluating 1D kernel with %d observations' % len(X)
kernel = stats.kde.gaussian_kde(X)
if Verbose:
print 'Evaluating 1D KDE with %d bins' % N
P = kernel(bins)
return (bins, P)
def difference_CI_bootstrap_wrapper(data, **boot_kwargs):
"""Given parsed data from single ulabel, return difference CIs.
data : same format as bootstrap_main_effect expects
Will calculate the following statistics:
means : mean of each condition, across draws
CIs : confidence intervals on each condition
mean_difference : mean difference between conditions
difference_CI : confidence interval on difference between conditions
p : two-tailed p-value of 'no difference'
Returns:
dict of those statistics
"""
# Yields a 1000 x 2 x N_trials matrix:
# 1000 draws from the original data, under both conditions.
bh = bootstrap_main_effect(data, meth=keep, **boot_kwargs)
# Find the distribution of means of each draw, across trials
# This is 1000 x 2, one for each condition
# hist(means_of_all_draws) shows the comparison across conditions
means_of_all_draws = bh.mean(axis=2)
# Confidence intervals across the draw means for each condition
condition_CIs = np.array([
mlab.prctile(dist, (2.5, 97.5)) for dist in means_of_all_draws.T])
# Means of each ulabel (centers of the CIs, basically)
condition_means = means_of_all_draws.mean(axis=0)
# Now the CI on the *difference between conditions*
difference_of_conditions = np.diff(means_of_all_draws).flatten()
difference_CI = mlab.prctile(difference_of_conditions, (2.5, 97.5))
# p-value of 0. in the difference distribution
cdf_at_value = np.sum(difference_of_conditions < 0.) / \
float(len(difference_of_conditions))
p_at_value = 2 * np.min([cdf_at_value, 1 - cdf_at_value])
# Should probably floor the p-value at 1/n_boots
return {'p' : p_at_value,
'means' : condition_means, 'CIs': condition_CIs,
'mean_difference': difference_of_conditions.mean(),
'difference_CI' : difference_CI}
def get_sample_percentiles(self, percents):
'It returns the percentiles given a percent list'
if not self._sample:
raise ValueError('No data to calculate percentiles')
vect = numpy.ravel(self.sample)
percentiles = mlab.prctile(vect, percents)
return list(percentiles)
def percentile_box_plot(ax, data, indexer=None, box_top=75,
box_bottom=25,whisker_top=98,whisker_bottom=2):
if indexer is None:
indexed_data = zip(range(1,len(data)+1), data)
else:
indexed_data = [(indexer(datum), datum) for datum in data]
for index, x in indexed_data:
if whisker_top != None and whisker_bottom != None:
bp = boxplotter(*(prctile(x,(50,box_top,box_bottom,whisker_top,whisker_bottom))))
bp.draw_on(ax, index, data=x)
elif whisker_top == None and whisker_bottom == None:
bp = boxplotter(*(prctile(x,(50,box_top,box_bottom))))
bp.draw_on(ax, index)
else:
raise Exception("Just one whisker? That's silly.")
def identify_outliers(test, chains, x):
"""
Determine which chains have converged on a local maximum much lower than
the maximum likelihood.
*test* is the name of the test to use (one of IQR, Grubbs, Mahal or none).
*chains* is a set of log likelihood values of shape (chain len, num chains)
*x* is the current population of shape (num vars, num chains)
See :module:`outliers` for details.
"""
# Determine the mean log density of the active chains
v = mean(chains, axis=0)
# Check whether any of these active chains are outlier chains
test = test.lower()
if test == 'iqr':
# Derive the upper and lower quartile of the chain averages
Q1,Q3 = prctile(v,[25,75])
# Derive the Inter Quartile Range (IQR)
IQR = Q3 - Q1
# See whether there are any outlier chains
outliers = where(v < Q1 - 2*IQR)[0]
elif test == 'grubbs':
# Compute zscore for chain averages
zscore = (mean(v) - v) / std(v, ddof=1)
# Determine t-value of one-sided interval
N = len(v)
t2 = tinv(1 - 0.01/N,N-2)**2; # 95% interval
# Determine the critical value
Gcrit = ((N - 1)/sqrt(N)) * sqrt(t2/(N-2 + t2))
# Then check against this
outliers = where(zscore > Gcrit)[0]
elif test == 'mahal':
# Use the Mahalanobis distance to find outliers in the population
alpha = 0.01
Npop, Nvar = x.shape
Gcrit = ACR(Nvar,Npop-1,alpha)
#print "alpha",alpha,"Nvar",Nvar,"Npop",Npop,"Gcrit",Gcrit
# Find which chain has minimum log_density
minidx = argmin(v)
# Then check the Mahalanobis distance of the current point to other chains
d1 = mahalanobis(x[minidx,:], x[minidx!=arange(Npop),:])
#print "d1",d1,"minidx",minidx
# and see if it is an outlier
outliers = [minidx] if d1 > Gcrit else []
elif test == 'none':
outliers = []
else:
raise ValueError("Unknown outlier test "+test)
return outliers
def do_kde(X, range=None, N=256):
if range is None:
prc = prctile(X.ravel())
a = prc[0]
b = prc[4]
else:
a, b = range
bins = linspace(a, b, N)
kernel = kde.gaussian_kde(X.ravel())
return bins, kernel(bins)
def ThresCal(self):
SampNum = 1000
KL = []
for i in range(0, SampNum):
x = chain(self.mu_0, self.P, self.n)
mu = np.reshape(self.mu, (self.N, self.N))
KL.append(KL_est(
x, mu)) # Get the actual relative entropy (K-L divergence)
eta = prctile(self.KL, 100 * (1 - self.beta))
return eta
def ThresCal(self):
SampNum = 1000
self.KL = []
for i in range(0, SampNum):
x = chain(self.mu_0, self.P, self.n)
mu = np.reshape(self.mu, (self.N, self.N))
self.KL.append(KL_est(x, mu)) # Get the actual relative entropy (K-L divergence)
self.eta = prctile(self.KL, 100 * (1 - self.beta))
KL = self.KL
eta = self.eta
return KL, eta
def modeifyer(times, fluxes, window=500, p=20, minpoints=10):
"""Uses percentile p of points around each datapoint "flux",
being within time window to detrend fluxes. Returns
corrected fluxes. For now done with a slow loop..."""
detrend = fluxes.copy()
for i in range(len(times)):
near_fluxes = fluxes[where(
(times < times[i] + window / 2) * (times > times[i] - window / 2))]
trend = prctile(near_fluxes, p)
detrend[i] = fluxes[i] - trend
return detrend
def prctile(self, p = (2.5, 97.5)):
''' Returns the standard percentiles of the bootstrapped statistic.
Arguments
---------
perc :
A sequence of percentile values or a scalar
'''
from matplotlib.mlab import prctile
return prctile(self.dist, p = p)
def modeifyer(times,fluxes,window=500,p=20,minpoints=10):
"""Uses percentile p of points around each datapoint "flux",
being within time window to detrend fluxes. Returns
corrected fluxes. For now done with a slow loop..."""
detrend = fluxes.copy()
for i in range(len(times)):
near_fluxes = fluxes[where((times<times[i]+window/2)*(
times>times[i]-window/2))]
trend = prctile(near_fluxes,p)
detrend[i] = fluxes[i] - trend
return detrend
def _calculate_percentiles(numbers, percents):
'It calculates the percentiles for some numbers'
#we need a numpy array
if 'any' not in dir(numbers):
numbers = numpy.ravel(numbers)
if not numbers.any():
raise ValueError('No data to calculate percentiles')
mlab = sys.modules['matplotlib.mlab']
percentiles = mlab.prctile(numbers, percents)
return list(percentiles)
def bootstrap_regress(x, y, n_boot=1000):
from matplotlib import mlab
x = np.asarray(x)
y = np.asarray(y)
m_l, b_l = [], []
for n in range(n_boot):
msk = np.random.randint(0, len(x), size=len(x))
m, b, rval, pval, stderr = scipy.stats.stats.linregress(x[msk], y[msk])
m_l.append(m)
b_l.append(b)
res = {
'slope_m': np.mean(m_l),
'slope_l': mlab.prctile(m_l, p=2.5),
'slope_h': mlab.prctile(m_l, p=97.5),
'intercept_m': np.mean(b_l),
'intercept_l': mlab.prctile(b_l, p=2.5),
'intercept_h': mlab.prctile(b_l, p=97.5),
}
return res
def plot_kde(X, a=None, b=None, N=256, Title=None, Label=None):
if a == None:
prc = prctile(X.ravel())
a = prc[0]
b = prc[4]
if Title is None:
Title = 'Kernel Density Function'
bins = linspace(a, b, N)
kernel = kde.gaussian_kde(X.ravel())
plot(bins, kernel(bins))
ylabel('PDF')
title(Title)
def bootstrap_regress(x, y, n_boot=1000):
from matplotlib import mlab
x = np.asarray(x)
y = np.asarray(y)
m_l, b_l = [], []
for n in range(n_boot):
msk = np.random.randint(0, len(x), size=len(x))
m, b, rval, pval, stderr = scipy.stats.stats.linregress(x[msk], y[msk])
m_l.append(m)
b_l.append(b)
res = {
"slope_m": np.mean(m_l),
"slope_l": mlab.prctile(m_l, p=2.5),
"slope_h": mlab.prctile(m_l, p=97.5),
"intercept_m": np.mean(b_l),
"intercept_l": mlab.prctile(b_l, p=2.5),
"intercept_h": mlab.prctile(b_l, p=97.5),
}
return res
def test_prctile():
# test odd lengths
x=[1,2,3]
assert mlab.prctile(x,50)==np.median(x)
# test even lengths
x=[1,2,3,4]
assert mlab.prctile(x,50)==np.median(x)
# derived from email sent by jason-sage to MPL-user on 20090914
ob1=[1,1,2,2,1,2,4,3,2,2,2,3,4,5,6,7,8,9,7,6,4,5,5]
p = [0, 75, 100]
expected = [1, 5.5, 9]
# test vectorized
actual = mlab.prctile(ob1,p)
assert np.allclose( expected, actual )
# test scalar
for pi, expectedi in zip(p,expected):
actuali = mlab.prctile(ob1,pi)
assert np.allclose( expectedi, actuali )
def test_prctile():
# test odd lengths
x = [1, 2, 3]
assert mlab.prctile(x, 50) == np.median(x)
# test even lengths
x = [1, 2, 3, 4]
assert mlab.prctile(x, 50) == np.median(x)
# derived from email sent by jason-sage to MPL-user on 20090914
ob1 = [1, 1, 2, 2, 1, 2, 4, 3, 2, 2, 2, 3, 4, 5, 6, 7, 8, 9, 7, 6, 4, 5, 5]
p = [0, 75, 100]
expected = [1, 5.5, 9]
# test vectorized
actual = mlab.prctile(ob1, p)
assert np.allclose(expected, actual)
# test scalar
for pi, expectedi in zip(p, expected):
actuali = mlab.prctile(ob1, pi)
assert np.allclose(expectedi, actuali)
def main_chain(img_name, template_name, blursize=5, cliplim=3.0, gridsize=8):
# FIXME what if foscam moves, then blind offset-from-template method will not work robustly, will it?
# apply blurring and CLAHE to small (skinny garage door) roi
img, final, xywh_template, topleft_sgd, botright_sgd = blurred_histogram_equalization(
img_name, template_name, blursize=5, cliplim=3.0, gridsize=8)
# histogram of skinny garage door subset after image processing
colors = ('b', )
fig, ax = plt.subplots(figsize=(12, 8))
# ith-channel of skinny garage door (roi1) after image processing
i = 0
c = colors[i]
sgd = final[topleft_sgd[1]:botright_sgd[1],
topleft_sgd[0]:botright_sgd[0]][:, :, i] # i = 0 for Luminance
intensity_bins = range(0, 256)
n, bins, patches = ax.hist([sgd],
intensity_bins,
normed=1,
color=c,
histtype='step',
cumulative=True,
label='Color: ' + c)
# FIXME we may not always want histogram plot (maybe just during "gather")
# TODO with each image file, always ratchet up a running sum type of histogram OR db each for future summing
# TODO always put percentiles (10th, 20th, 30th, ... 90th) into db table!?
# percentiles
percs = mlab.prctile([sgd], p=np.arange(10.0, 100.0, 10.0))
print percs
# tidy up the figure
ax.grid(True)
#ax.legend(loc='right')
#ax.set_title('Cumulative Step Histograms')
ax.set_title('Cumulative Step Histogram, Blue Channel, %s' % img_name)
ax.set_xlabel('Pixel [intensity?]')
ax.set_ylabel('Likelihood of Occurrence')
plt.xlim([0, 256])
# save cumulative histogram figure as _chist.jpg
outname = img_name.replace('.jpg', '_chist.jpg')
plt.savefig(outname)
print 'open -a Firefox file://%s' % outname
return img, final, xywh_template
def simple_bootstrap(data, n_boots=1000, min_bucket=20):
if len(data) < min_bucket:
raise BootstrapError("too few samples")
res = []
data = np.asarray(data)
for boot in range(n_boots):
idxs = np.random.randint(0, len(data), len(data))
draw = data[idxs]
res.append(np.mean(draw))
res = np.asarray(res)
CI = mlab.prctile(res, (2.5, 97.5))
return res, res.mean(), CI
def HoeffdingRuleMarkov(beta, rho, G, H, W, Chi, FlowNum):
"""
Estimate the K-L divergence and the threshold by use of weak convergence
----------------
beta: the false alarm rate
mu: the stationary distribution
G: the gradient
H: the Hessian
Sigma: the covariance matrix
W: a sample path of the Gaussian empirical measure
Chi: a sample path of the "Chi-Square" estimation
FlowNum: the number of flows
----------------
"""
_, SampNum, N = W.shape # Here, N equals the number of states in the new chain Z
# Estimate K-L divergence using 2nd-order Taylor expansion
KL_1 = []
for j in range(0, SampNum):
t = (1.0 / sqrt(FlowNum)) * np.dot(G, W[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(W[0, j, :], H), W[0, j, :])
# print t.tolist()
# break
KL_1.append(np.array(t.real)[0])
# Get the threshold
eta1 = prctile(KL_1, 100 * (1 - beta))
KL_2 = [Chi[idx] / (2 * FlowNum) for idx in xrange(len(Chi))]
# Using the simplified formula
# eta2 = 1.0 / (2 * FlowNum) * rho * chi2.ppf(1 - beta, N)
eta2 = prctile(KL_2, 100 * (1 - beta))
# print N
# print(KL)
# assert(1 == 2)
return KL_1, KL_2, eta1, eta2
def bootstrapMedian(data, N=5000):
'''Bootstraper to refine estimate of a percentile from data
N = number of iterations for the bootstrapping
M = number of data points
output = MU.bootStrapper(data, 50, 10000)
'''
import numpy as np
import matplotlib.mlab as mlab
M = len(data)
percentile = 50
estimate = np.array([])
for k in range(N):
bsIndex = np.random.random_integers(0,M-1,M)
bsData = data[bsIndex]
tmp = mlab.prctile(bsData, percentile)
estimate = np.hstack((estimate, tmp))
CI = mlab.prctile(estimate, [2.5,97.5])
med = np.mean(estimate)
return med, CI, estimate
def calc_kde2d(x_values,
y_values,
x_range=None,
y_range=None,
Nx=256,
Ny=256,
npz=None,
Verbose=True,
name=None):
if Verbose:
if name != None:
print "[] ", name
print "Starting the 2D kernel density estimation with %d data points..."\
%len(x_values)
kernel = stats.kde.gaussian_kde(_cat(x_values, y_values))
if x_range == None:
prc = prctile(x_values)
x_range = [prc[0], prc[4]]
if y_range == None:
y_range = x_range
x_bins = linspace(x_range[0], x_range[1], Nx)
y_bins = linspace(y_range[0], y_range[1], Ny)
if Verbose:
print "Evaluating 2D kernel on grid with (Nx,Ny)=(%d,%d) ..." % (Nx,
Ny)
X, Y = meshgrid(x_bins, y_bins) # each has shape (Ny,Nx)
P = kernel(_cat(X, Y)) # shape is (Ny*Nx)
P = reshape(P, X.shape)
if Verbose:
print "X, Y, P shapes: ", X.shape, Y.shape, P.shape
# Save to file
# ------------
if npz != None:
print "Saving 2D KDE to file <" + npz + "> ..."
savez(npz, pdf=P, x_bins=x_bins, y_bins=y_bins)
return (x_bins, y_bins, P)
def test_Median(self):
'test median'
# Wiki: If there is an even number of observations, then there is no single middle value; the median is then usually defined to be the
# mean of the two middle values.[1][2]
lol = [
# [], # fails
[1.2],
[1.0, 2.0], # Get 1.5 (matplotlib 1.0.1 or 2.0 (matplotlib 0.99.3)
[1.0, 2.0, 4.0],
]
expectedMedianList = [ 1.2, 1.5, 2.0] # matplotlib 1.0.1
expectedMedianListOldMatplotlib = [ 1.2, 2.0, 2.0] # matplotlib 0.99.3
for i,floatList in enumerate(lol):
ml = mlab.prctile(floatList,[50])
nTdebug("Found: %s and expected (by new matplotlib): %s" % (ml[0], expectedMedianList[i]))
if ml[0] != expectedMedianList[i]:
self.assertEqual(ml[0], expectedMedianListOldMatplotlib[i])
def boxpoints(d, outlier_distance=1.5):
# implementation pretty much the same as matplotlib axes.boxplot
# get median and quartiles
q1, med, q3 = mlab.prctile(d,[25,50,75])
# min(data), max(data)
iq = q3 - q1
hi_val = q3 + outlier_distance*iq
lo_val = q1 - outlier_distance*iq
print iq, q1, q3, '---', hi_val, lo_val
# print (d > hi_val)
# print (d < lo_val)
outliers = r_[d[d>hi_val], d[d<lo_val]]
# print 'outliers', outliers
inliers = list(set(data)-set(outliers))
# print 'inliers', inliers
min_without_outliers = min(inliers)
max_without_outliers = max(inliers)
return outliers, min_without_outliers, q1, med, q3, max_without_outliers
def plot_transparent_histogram(arr,
ax,
frame_width,
frame_height,
upper_prctile_clim=95,
cmap=plt.cm.gray_r):
"""Imshow a histogram with zero values transparent
H_tip : single 2d array to plot. Should be non-negative
frame_width, frame_height : passed to imshow to get the data limits right
All zero bins will be transparent. The image color limits will be
set to the 95th percentile of non-zero values.
"""
# Determine the transparent threshold and upper clim
vals = arr.flatten()
vals = vals[vals > 0]
transparent_threshold = vals.min()
clim_upper = prctile(vals, upper_prctile_clim)
# Mask the data to make zero bins transparent
# We use .99 to avoid floating point comparison problems
masked_data = np.ma.masked_where(arr < transparent_threshold * .99, arr)
# Plot
im = my.plot.imshow(
masked_data,
ax=ax,
xd_range=(0, frame_width),
yd_range=(0, frame_height),
axis_call='image',
cmap=cmap,
skip_coerce=True,
)
# Set the clim to go from 0 to upper
im.set_clim((0, clim_upper))
return im
def HoeffdingRuleMarkovRobust(beta, G_1, H_1, U_1, G_2, H_2, U_2, G_3, H_3,
U_3, FlowNum):
"""
Estimate the K-L divergence and the threshold by use of weak convergence
----------------
beta: the false alarm rate
G: the gradient
H: the Hessian
U: a sample path of the Gaussian empirical measure
FlowNum: the number of flows
----------------
"""
_, SampNum, _ = U_1.shape
# Estimate K-L divergence using 2nd-order Taylor expansion
KL = []
for j in range(0, SampNum):
t_1 = (1.0 / sqrt(FlowNum)) * np.dot(G_1, U_1[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(U_1[0, j, :], H_1), U_1[0, j, :])
t_2 = (1.0 / sqrt(FlowNum)) * np.dot(G_2, U_2[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(U_2[0, j, :], H_2), U_2[0, j, :])
t_3 = (1.0 / sqrt(FlowNum)) * np.dot(G_3, U_3[0, j, :]) + \
(1.0 / 2) * (1.0 / FlowNum) * \
np.dot(np.dot(U_3[0, j, :], H_3), U_3[0, j, :])
t1 = np.array(t_1.real)[0]
t2 = np.array(t_2.real)[0]
t3 = np.array(t_3.real)[0]
# print t.tolist()
# break
KL.append(min([t1, t2, t3]))
eta = prctile(KL, 100 * (1 - beta))
# print(KL)
# assert(1 == 2)
return eta
def bootstrapped_intercluster_mahalanobis(cluster1,
cluster2,
n_boots=1000,
fix_covariances=True):
"""Bootstrap the intercluster distance.
Returns:
m - The mean distance
CI - 95% confidence interval on the distance
distances - an array of the distances measured on each boot
"""
d_l = []
# Determine the covariance matrices, or recalculate each time
if fix_covariances:
icov1 = np.linalg.inv(np.cov(cluster1, rowvar=0))
icov2 = np.linalg.inv(np.cov(cluster2, rowvar=0))
else:
icov1, icov2 = None, None
# Bootstrap
for n_boot in range(n_boots):
# Draw
idxs1 = np.random.randint(0, len(cluster1), len(cluster1))
idxs2 = np.random.randint(0, len(cluster2), len(cluster2))
# Calculate and store
d_l.append(
intercluster_mahalanobis(cluster1[idxs1], cluster2[idxs2], icov1,
icov2))
# Statistics
d_a = np.asarray(d_l)
m = np.mean(d_a)
CI = mlab.prctile(d_a, (2.5, 97.5))
return m, CI, d_a
datas = [np.recfromtxt(f, names = True, case_sensitive = True) for f in files]
data = np.ma.concatenate(datas)
desired_unit = dict(O3 = 'ppb', GMAO_TEMP = 'K', PRESS = 'hPa', TEMP = 'K')
unit_factor = {'ppt': 1e12, 'ppb': 1e9}
pfile.createDimension('time', data.shape[0])
for ki, key in enumerate(data.dtype.names):
typecode = data[key].dtype.char
if typecode not in ('c', 'S'):
unit = desired_unit.get(key, 'ppt')
factor = unit_factor.get(unit, 1)
values = np.ma.masked_values(data[key], -1000) * factor
else:
unit = 'unknown'
values = data[key]
pfile.createVariable(key, typecode, dimensions = ('time',), units = unit, values = values)
if __name__ == '__main__':
import sys
bfile1 = flightlogs(sys.argv[1:])
from matplotlib.mlab import prctile
for label, key in [('O3', 'O3[:]'), ('NO2', 'NO2[:]')]:
bvar = eval(key, None, bfile1.variables)
b2var = eval(key, None, bfile1.variables)
assert((bvar == b2var).all())
print('\n%s (BASE: %6.2f)' % (label, bvar.mean()), file = sys.stdout)
print('\n BASE:', sep = '', file = sys.stdout)
prctile(bvar, np.ma.arange(.1, 1., .1)* 100).tofile(sys.stdout, sep = ', ', format = '%6.2f')
print('', file = sys.stdout)
def read_state_trajectories():
global state_array, NUM_DIM, NUM_STATES, TRAJ_LEN
state_trajs = []
to_put = []
rewards = []
trajs = open("state_trajectories.dat", 'r')
if trajs:
lines = trajs.readlines()
for l in lines:
s = l.split('\t')
num_steps = len(s) -1
if( len(s) > 3):
to_put = [int(s[x]) for x in range(num_steps)]
state_trajs.append(to_put)
if len(to_put) > TRAJ_LEN:
TRAJ_LEN = len(to_put)
else:
rewards.append(float(s[1]))
trajs.close()
num_traj = len(state_trajs)
for ct in range(num_traj):
last = state_trajs[ct][-1]
curr_len = len(state_trajs[ct])
for ti in np.linspace(curr_len, TRAJ_LEN-1, TRAJ_LEN-curr_len):
state_trajs[ct].append(last)
state_trajs = np.array(state_trajs)
traj_len = len(state_trajs[0])
print "num_traj: ", num_traj , " traj_len: ", traj_len, " reward: ", np.average(rewards)
"""
fig = figure(3)
fig.add_subplot(111, aspect='equal')
# create a hexbin map now for each trajectory
for i in range(len(state_trajs)):
curr_traj = np.array([ [state_array[x,0], state_array[x,1]] for x in state_trajs[i] ] )
clf()
scatter( curr_traj[:,0], curr_traj[:,1], marker='o', c='y', s= 25, alpha=0.7)
#hexbin(curr_traj[:,0], curr_traj[:,1], gridsize=10, cmap=cm.get_cmap('Jet'), alpha=0.9, mincnt=1)
fig.savefig("movie/"+str(i)+".png")
"""
"""
if NUM_DIM==2:
fig = figure(1)
ax = fig.add_subplot(111, aspect='equal')
for i in range(len(state_trajs)):
curr_traj = np.array([ [state_array[x,i] for i in range(NUM_DIM)] for x in state_trajs[i] ] )
plot(curr_traj[:,0], curr_traj[:,1], 'b-', lw=0.5, alpha=0.2)
circle = Circle( (curr_traj[0,0], curr_traj[0,1]), 0.01, fc='red', alpha = 0.4)
ax.add_patch(circle)
circle = Circle( (curr_traj[traj_len-1,0], curr_traj[traj_len-1,1]), 0.01, fc='green', alpha = 0.4)
ax.add_patch(circle)
"""
fig = figure(2)
state_traj_x = []
state_traj_y = []
for i in range(num_traj):
curr_traj = np.array([ [state_array[x,j] for j in range(NUM_DIM)] for x in state_trajs[i] ] )
tmp = np.array([state_array[x,0] for x in state_trajs[i]])
state_traj_x.append(tmp)
if NUM_DIM == 2:
tmp = np.array([state_array[x,1] for x in state_trajs[i]])
state_traj_y.append(tmp)
#subplot(111)
#plot(curr_traj[:,0], 'b-', lw=0.5, alpha=0.10)
#subplot(212)
#plot(curr_traj[:,1], 'ro', lw=0.5, alpha=0.05)
state_traj_x = np.array(state_traj_x)
state_traj_y = np.array(state_traj_y)
print state_traj_x.shape, state_traj_y.shape
state_traj_x_percentile_10 = np.array([mlab.prctile(state_traj_x[:,i],p=10) for i in range(TRAJ_LEN)])
state_traj_x_percentile_50 = np.array([mlab.prctile(state_traj_x[:,i],p=50) for i in range(TRAJ_LEN)])
state_traj_x_percentile_90 = np.array([mlab.prctile(state_traj_x[:,i],p=90) for i in range(TRAJ_LEN)])
state_traj_x_percentile = np.array([state_traj_x_percentile_10, state_traj_x_percentile_90])
if NUM_DIM == 2:
state_traj_y_percentile_10 = np.array([mlab.prctile(state_traj_y[:,i],p=10) for i in range(TRAJ_LEN)])
state_traj_y_percentile_90 = np.array([mlab.prctile(state_traj_y[:,i],p=90) for i in range(TRAJ_LEN)])
state_traj_y_percentile = np.array([state_traj_y_percentile_10, state_traj_y_percentile_90])
subplot(211)
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), np.average(state_traj_x, axis=0), 'b-', label='mean')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_x_percentile_10, 'b--', label='10/90 percentile')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_x_percentile_90, 'b--')
legend()
grid()
ylabel('x (t)')
xlabel('t [s]')
axis('tight')
subplot(212)
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), np.average(state_traj_y, axis=0), 'b-', label='mean')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_y_percentile_10, 'b--', label='10/90 percentile')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_y_percentile_90, 'b--')
legend()
grid()
ylabel('y (t)')
xlabel('t [s]')
axis('tight')
elif NUM_DIM==1:
subplot(111)
#plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), np.average(state_traj_x, axis=0), 'b-', label='mean')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_x_percentile_10, 'b--', label='10/50/90 percentile')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_x_percentile_90, 'b--')
plot( holding_time*np.linspace(0,TRAJ_LEN,num=TRAJ_LEN), state_traj_x_percentile_50, 'b--')
legend()
grid()
xlabel('t [s]')
axis('tight')
def imshow(arr, x=None, ax=None, vmin=None, vmax=None, percentile=True,
strip=False, features=None, conf=0.95, line_kwargs=None,
sort_by=None, fill_kwargs=None, figsize=(5, 12),
width_ratios=(4, 1), height_ratios=(4, 1),
subplot_params=dict(wspace=0.1, hspace=0.1), imshow_kwargs=None):
"""
Parameters
----------
arr : array-like
x : 1D array
X values to use. If None, use range(arr.shape[1])
ax : matplotlib.Axes
If not None, then only plot the array on the provided axes. This will
ignore any additional arguments provided that apply to figure-level
configuration or to the average line plot. For example, `figsize`,
`width_ratios`, `height_ratios`, `subplot_params`, `line_kwargs`, and
`fill_kwargs` will be ignored.
vmin, vmax : float
percentile : bool
If True, then treat values for `vmin` and `vmax` as percentiles rather
than absolute values.
strip : bool
Include a strip plot alongside the array
features : pybedtools.BedTool or string filename
Features used to construct the array
sort_by : array-like
Use the provided array to sort the array (e.g., expression). This
array is argsorted to get the proper order.
line_kwargs, fill_kwargs : dict
Passed directly to `ci_plot`.
figsize : tuple
(Width, height) of the figure to create.
"""
if ax is None:
fig = new_shell(
figsize=figsize,
strip=strip,
subplot_params=subplot_params,
width_ratios=width_ratios,
height_ratios=height_ratios)
if x is None:
x = np.arange(arr.shape[1])
if percentile:
if vmin is None:
vmin = arr.min()
else:
vmin = mlab.prctile(arr.ravel(), vmin)
if vmax is None:
vmax = arr.max()
else:
vmax = mlab.prctile(arr.ravel(), vmax)
else:
if vmin is None:
vmin = arr.min()
if vmax is None:
vmax = arr.max()
if imshow_kwargs is None:
imshow_kwargs = {}
cmap = colormap_adjust.smart_colormap(vmin, vmax)
if sort_by is not None:
ind = np.argsort(sort_by)
else:
ind = np.arange(arr.shape[0])
if ax is None:
array_ax = fig.array_axes
else:
array_ax = ax
mappable = array_ax.imshow(
arr[ind, :],
aspect='auto',
cmap=cmap,
vmin=vmin,
vmax=vmax,
origin='lower',
extent=(x.min(), x.max(), 0, arr.shape[0]),
**imshow_kwargs
)
if ax is None:
plt.colorbar(mappable, fig.cax)
ci_plot(
x,
arr,
ax=fig.line_axes,
line_kwargs=line_kwargs,
fill_kwargs=fill_kwargs,
)
return fig
else:
return ax.figure
def input_ip_plots(iparr, inputarr, diffed, x, sort_ind,
prefix=None, limits1=(None, None), limits2=(None, None),
hlines=None, vlines=None):
"""
All-in-one plotting function to make a 5-panel figure.
Panels are IP, input, and diffed; plus 2 line plots showing averages.
:param iparr, inputarr: NumPy arrays constructed by a genomic_signal object
:param diffed: Difference of `iparr` and `inputarr`, but can be some other
transformation.
:param x: Extent to use -- for TSSs, maybe something like
np.linspace(-1000, 1000, bins), or for just bin IDs, something like
`np.arange(bins)`.
:param sort_ind: row order for each of the 3 panels -- usually interesting
to use `clustered_sortind` or `tip_zscores`
:param prefix: Used to prefix plot titles with '%(prefix)s IP", etc
:param limits1: Tuple passed to the Normalize function for IP and input.
:param limits2: Tuple passed tot he Normalize function for the diffed array
:param hlines: List of (position, kwarg) tuples for plotting horizontal
lines. Kwargs are passed directly to axhline. Useful for delimiting
clusters, if you used `clustered_sortind` and have both `row_order` and
`breaks`.
:param vlines: List of (position, kwargs) tuples. A vertical line will be
plotted at each position using kwargs.
"""
# global min and max
gmin = min(iparr.min(), inputarr.min())
gmax = max(iparr.max(), inputarr.max())
fig = plt.figure(figsize=(10, 10))
# 3 arrays, 2 line plots, a gene strip, and 2 colorbars. Plots share the
# axes that make sense
#
# 3 arrays
ax1 = plt.subplot2grid(
(9, 9), (0, 0), colspan=3, rowspan=6)
ax2 = plt.subplot2grid(
(9, 9), (0, 3), colspan=3, rowspan=6, sharex=ax1, sharey=ax1)
ax3 = plt.subplot2grid(
(9, 9), (0, 6), colspan=3, rowspan=6, sharex=ax1, sharey=ax1)
# 2 line plots
ax4 = plt.subplot2grid((9, 9), (6, 3), colspan=3, rowspan=3, sharex=ax1)
ax5 = plt.subplot2grid((9, 9), (6, 6), colspan=3, rowspan=3, sharex=ax1)
# 2 colorbars
cax1 = plt.Axes(fig, rect=(0.05, 0.25, 0.25, 0.025))
cax2 = plt.Axes(fig, rect=(0.05, 0.15, 0.25, 0.025))
# For nice imshow axes
extent = (min(x), max(x), 0, diffed.shape[0])
cm = matplotlib.cm.gist_gray
cm.set_bad('k')
cm.set_over('r')
cm.set_under('b')
limits1 = list(limits1)
limits2 = list(limits2)
all_base = np.column_stack((iparr.ravel(), inputarr.ravel())).ravel()
if limits1[0] is None:
limits1[0] = mlab.prctile(
all_base, 1. / all_base.size)
if limits1[1] is None:
limits1[1] = mlab.prctile(
all_base, 100 - 1. / all_base.size)
if limits2[0] is None:
limits2[0] = mlab.prctile(
diffed.ravel(), 1. / all_base.size)
if limits2[1] is None:
limits2[1] = mlab.prctile(
diffed.ravel(), 100 - 1. / all_base.size)
del all_base
imshow_kwargs = dict(
interpolation='nearest',
aspect='auto',
cmap=cm,
norm=matplotlib.colors.Normalize(*limits1),
extent=extent,
origin='lower')
# modify kwargs for diffed (by changing the normalization)
diffed_kwargs = imshow_kwargs.copy()
diffed_kwargs['norm'] = matplotlib.colors.Normalize(*limits2)
# IP
mappable1 = ax1.imshow(iparr[sort_ind, :], **imshow_kwargs)
# input
mappable2 = ax2.imshow(inputarr[sort_ind, :], **imshow_kwargs)
# diffed
mappable3 = ax3.imshow((diffed)[sort_ind, :], **diffed_kwargs)
# IP and input line plot with vertical line
ax4.plot(x, inputarr.mean(axis=0), color='k', linestyle='--',
label='input')
ax4.plot(x, iparr.mean(axis=0), color='k', label='ip')
ax4.axvline(0, color='k', linestyle=':')
# Diffed line plot with vertical line
ax5.plot(x, diffed.mean(axis=0), 'k', label='enrichment')
ax5.axvline(0, color='k', linestyle=':')
# Colorbars
cbar1 = fig.colorbar(mappable1, cax1, orientation='horizontal')
cbar2 = fig.colorbar(mappable3, cax2, orientation='horizontal')
fig.add_axes(cax1)
fig.add_axes(cax2)
# labeling...
ax1.set_ylabel('features')
plt.setp(ax2.get_yticklabels(), visible=False)
plt.setp(ax3.get_yticklabels(), visible=False)
ax4.set_xlabel('bp')
ax4.set_ylabel('mean reads per million mapped reads')
ax5.set_xlabel('bp')
cax1.set_xlabel('Reads per million mapped reads')
cax2.set_xlabel('Enrichment (RPMMR)')
if prefix is None:
prefix = ""
ax1.set_title('%s IP' % prefix)
ax2.set_title('%s input' % prefix)
ax3.set_title('Difference')
# diffed line plot should have y ax on right
ax5.yaxis.set_ticks_position('right')
ax5.yaxis.set_label_position('right')
ax5.set_ylabel('enriched reads per million mapped reads')
# Legends
ax4.legend(loc='best', frameon=False)
ax5.legend(loc='best', frameon=False)
# Make sure everybody snaps to xmin/xmax
for ax in [ax1, ax2, ax3, ax4, ax5]:
ax.axis(xmin=extent[0], xmax=extent[1])
if not hlines:
hlines = []
if not vlines:
vlines = []
for ax in [ax1, ax2, ax3]:
for pos, kwargs in hlines:
ax.axhline(pos, **kwargs)
for pos, kwargs in vlines:
ax.axvline(pos, **kwargs)
fig.subplots_adjust(bottom=0.05, top=0.95, hspace=0.75, wspace=0.9)
return fig
def boxplot(x, notch=0, sym='b+', positions=None, widths=None):
"""Makes a box and whisker plot.
Adapted from matplotlib.axes 0.98.5.2
Modified such that the caps are set to the 10th and 90th
percentiles, and to have some control on the colors.
call signature::
boxplot(x, notch=0, sym='+', positions=None, widths=None)
Make a box and whisker plot for each column of *x* or each
vector in sequence *x*. The box extends from the lower to
upper quartile values of the data, with a line at the median.
The whiskers extend from the box to show the range of the
data. Flier points are those past the end of the whiskers.
- *notch* = 0 (default) produces a rectangular box plot.
- *notch* = 1 will produce a notched box plot
*sym* (default 'b+') is the default symbol for flier points.
Enter an empty string ('') if you don't want to show fliers.
*whis* (default 1.5) defines the length of the whiskers as
a function of the inner quartile range. They extend to the
most extreme data point within ( ``whis*(75%-25%)`` ) data range.
*positions* (default 1,2,...,n) sets the horizontal positions of
the boxes. The ticks and limits are automatically set to match
the positions.
*widths* is either a scalar or a vector and sets the width of
each box. The default is 0.5, or ``0.15*(distance between extreme
positions)`` if that is smaller.
*x* is an array or a sequence of vectors.
Returns a dictionary mapping each component of the boxplot
to a list of the :class:`matplotlib.lines.Line2D`
instances created.
Copyright (c) 2002-2009 John D. Hunter; All Rights Reserved
"""
whiskers, caps, boxes, medians, fliers = [], [], [], [], []
# convert x to a list of vectors
if hasattr(x, 'shape'):
if len(x.shape) == 1:
if hasattr(x[0], 'shape'):
x = list(x)
else:
x = [
x,
]
elif len(x.shape) == 2:
nr, nc = x.shape
if nr == 1:
x = [x]
elif nc == 1:
x = [x.ravel()]
else:
x = [x[:, i] for i in xrange(nc)]
else:
raise ValueError, "input x can have no more than 2 dimensions"
if not hasattr(x[0], '__len__'):
x = [x]
col = len(x)
# get some plot info
if positions is None:
positions = range(1, col + 1)
if widths is None:
distance = max(positions) - min(positions)
widths = min(0.15 * max(distance, 1.0), 0.5)
if isinstance(widths, float) or isinstance(widths, int):
widths = numpy.ones((col, ), float) * widths
# loop through columns, adding each to plot
for i, pos in enumerate(positions):
d = numpy.ravel(x[i])
row = len(d)
# get median and quartiles
wisk_lo, q1, med, q3, wisk_hi = mlab.prctile(d, [10, 25, 50, 75, 90])
# get high extreme
#iq = q3 - q1
#hi_val = q3 + whis*iq
#wisk_hi = numpy.compress( d <= hi_val , d )
#if len(wisk_hi) == 0:
#wisk_hi = q3
#else:
#wisk_hi = max(wisk_hi)
## get low extreme
#lo_val = q1 - whis*iq
#wisk_lo = numpy.compress( d >= lo_val, d )
#if len(wisk_lo) == 0:
#wisk_lo = q1
#else:
#wisk_lo = min(wisk_lo)
# get fliers - if we are showing them
flier_hi = []
flier_lo = []
flier_hi_x = []
flier_lo_x = []
if len(sym) != 0:
flier_hi = numpy.compress(d > wisk_hi, d)
flier_lo = numpy.compress(d < wisk_lo, d)
flier_hi_x = numpy.ones(flier_hi.shape[0]) * pos
flier_lo_x = numpy.ones(flier_lo.shape[0]) * pos
# get x locations for fliers, whisker, whisker cap and box sides
box_x_min = pos - widths[i] * 0.5
box_x_max = pos + widths[i] * 0.5
wisk_x = numpy.ones(2) * pos
cap_x_min = pos - widths[i] * 0.25
cap_x_max = pos + widths[i] * 0.25
cap_x = [cap_x_min, cap_x_max]
# get y location for median
med_y = [med, med]
# calculate 'regular' plot
if notch == 0:
# make our box vectors
box_x = [box_x_min, box_x_max, box_x_max, box_x_min, box_x_min]
box_y = [q1, q1, q3, q3, q1]
# make our median line vectors
med_x = [box_x_min, box_x_max]
# calculate 'notch' plot
else:
raise NotImplementedError
notch_max = med #+ 1.57*iq/numpy.sqrt(row)
notch_min = med #- 1.57*iq/numpy.sqrt(row)
if notch_max > q3:
notch_max = q3
if notch_min < q1:
notch_min = q1
# make our notched box vectors
box_x = [
box_x_min, box_x_max, box_x_max, cap_x_max, box_x_max,
box_x_max, box_x_min, box_x_min, cap_x_min, box_x_min,
box_x_min
]
box_y = [
q1, q1, notch_min, med, notch_max, q3, q3, notch_max, med,
notch_min, q1
]
# make our median line vectors
med_x = [cap_x_min, cap_x_max]
med_y = [med, med]
doplot = plt.plot
whiskers.extend(
doplot(wisk_x, [q1, wisk_lo], color=whiskerscolor, linestyle='--'))
whiskers.extend(
doplot(wisk_x, [q3, wisk_hi], color=whiskerscolor, linestyle='--'))
caps.extend(
doplot(cap_x, [wisk_hi, wisk_hi], color=capscolor, linestyle='-'))
caps.extend(
doplot(cap_x, [wisk_lo, wisk_lo], color=capscolor, linestyle='-'))
boxes.extend(doplot(box_x, box_y, color=boxescolor, linestyle='-'))
medians.extend(doplot(med_x, med_y, color=medianscolor, linestyle='-'))
fliers.extend(
doplot(flier_hi_x, flier_hi, sym, flier_lo_x, flier_lo, sym))
# fix our axes/ticks up a little
newlimits = min(positions) - 0.5, max(positions) + 0.5
plt.gca().set_xlim(newlimits)
plt.gca().set_xticks(positions)
return dict(whiskers=whiskers,
caps=caps,
boxes=boxes,
medians=medians,
fliers=fliers)
def calculate_limits(array_dict, method='global', percentiles=None, limit=()):
"""
Calculate limits for a group of arrays in a flexible manner.
Returns a dictionary of calculated (vmin, vmax), with the same keys as
`array_dict`.
Useful for plotting heatmaps of multiple datasets, and the vmin/vmax values
of the colormaps need to be matched across all (or a subset) of heatmaps.
Parameters
----------
array_dict : dict of np.arrays
method : {'global', 'independent', callable}
If method="global", then use the global min/max values across all
arrays in array_dict. If method="independent", then each array will
have its own min/max calcuated. If a callable, then it will be used to
group the keys of `array_dict`, and each group will have its own
group-wise min/max calculated.
limit: tuple, optional
Tuple of 2 scalars passed directly to matplotlib.mlab.prctile to
limit the calculation of the percentile.
percentiles : None or list
If not None, a list of (lower, upper) percentiles in the range [0,100].
"""
if percentiles is not None:
for percentile in percentiles:
if not 0 <= percentile <= 100:
raise ValueError("percentile (%s) not between [0, 100]")
if method == 'global':
all_arrays = np.concatenate(
[i.ravel() for i in array_dict.itervalues()]
)
if percentiles:
vmin = mlab.prctile(
all_arrays, percentiles[0], limit=limit)
vmax = mlab.prctile(
all_arrays, percentiles[1], limit=limit)
else:
vmin = all_arrays.min()
vmax = all_arrays.max()
d = dict([(i, (vmin, vmax)) for i in array_dict.keys()])
elif method == 'independent':
d = {}
for k, v in array_dict.iteritems():
d[k] = (v.min(), v.max())
elif hasattr(method, '__call__'):
d = {}
sorted_keys = sorted(array_dict.keys(), key=method)
for group, keys in groupby(sorted_keys, method):
keys = list(keys)
all_arrays = np.concatenate([array_dict[i] for i in keys])
if percentiles:
vmin = mlab.prctile(
all_arrays, percentiles[0], limit=limit)
vmax = mlab.prctile(
all_arrays, percentiles[1], limit=limit)
else:
vmin = all_arrays.min()
vmax = all_arrays.max()
for key in keys:
d[key] = (vmin, vmax)
return d
def boxplot(x, notch=0, sym='b+', positions=None, widths=None):
"""Makes a box and whisker plot.
Adapted from matplotlib.axes 0.98.5.2
Modified such that the caps are set to the 10th and 90th
percentiles, and to have some control on the colors.
call signature::
boxplot(x, notch=0, sym='+', positions=None, widths=None)
Make a box and whisker plot for each column of *x* or each
vector in sequence *x*. The box extends from the lower to
upper quartile values of the data, with a line at the median.
The whiskers extend from the box to show the range of the
data. Flier points are those past the end of the whiskers.
- *notch* = 0 (default) produces a rectangular box plot.
- *notch* = 1 will produce a notched box plot
*sym* (default 'b+') is the default symbol for flier points.
Enter an empty string ('') if you don't want to show fliers.
*whis* (default 1.5) defines the length of the whiskers as
a function of the inner quartile range. They extend to the
most extreme data point within ( ``whis*(75%-25%)`` ) data range.
*positions* (default 1,2,...,n) sets the horizontal positions of
the boxes. The ticks and limits are automatically set to match
the positions.
*widths* is either a scalar or a vector and sets the width of
each box. The default is 0.5, or ``0.15*(distance between extreme
positions)`` if that is smaller.
*x* is an array or a sequence of vectors.
Returns a dictionary mapping each component of the boxplot
to a list of the :class:`matplotlib.lines.Line2D`
instances created.
Copyright (c) 2002-2009 John D. Hunter; All Rights Reserved
"""
whiskers, caps, boxes, medians, fliers = [], [], [], [], []
# convert x to a list of vectors
if hasattr(x, 'shape'):
if len(x.shape) == 1:
if hasattr(x[0], 'shape'):
x = list(x)
else:
x = [x,]
elif len(x.shape) == 2:
nr, nc = x.shape
if nr == 1:
x = [x]
elif nc == 1:
x = [x.ravel()]
else:
x = [x[:,i] for i in xrange(nc)]
else:
raise ValueError, "input x can have no more than 2 dimensions"
if not hasattr(x[0], '__len__'):
x = [x]
col = len(x)
# get some plot info
if positions is None:
positions = range(1, col + 1)
if widths is None:
distance = max(positions) - min(positions)
widths = min(0.15*max(distance,1.0), 0.5)
if isinstance(widths, float) or isinstance(widths, int):
widths = np.ones((col,), float) * widths
# loop through columns, adding each to plot
for i,pos in enumerate(positions):
d = np.ravel(x[i])
row = len(d)
# get median and quartiles
wisk_lo, q1, med, q3, wisk_hi = mlab.prctile(d,[10,25,50,75,90])
# get high extreme
#iq = q3 - q1
#hi_val = q3 + whis*iq
#wisk_hi = np.compress( d <= hi_val , d )
#if len(wisk_hi) == 0:
#wisk_hi = q3
#else:
#wisk_hi = max(wisk_hi)
## get low extreme
#lo_val = q1 - whis*iq
#wisk_lo = np.compress( d >= lo_val, d )
#if len(wisk_lo) == 0:
#wisk_lo = q1
#else:
#wisk_lo = min(wisk_lo)
# get fliers - if we are showing them
flier_hi = []
flier_lo = []
flier_hi_x = []
flier_lo_x = []
if len(sym) != 0:
flier_hi = np.compress( d > wisk_hi, d )
flier_lo = np.compress( d < wisk_lo, d )
flier_hi_x = np.ones(flier_hi.shape[0]) * pos
flier_lo_x = np.ones(flier_lo.shape[0]) * pos
# get x locations for fliers, whisker, whisker cap and box sides
box_x_min = pos - widths[i] * 0.5
box_x_max = pos + widths[i] * 0.5
wisk_x = np.ones(2) * pos
cap_x_min = pos - widths[i] * 0.25
cap_x_max = pos + widths[i] * 0.25
cap_x = [cap_x_min, cap_x_max]
# get y location for median
med_y = [med, med]
# calculate 'regular' plot
if notch == 0:
# make our box vectors
box_x = [box_x_min, box_x_max, box_x_max, box_x_min, box_x_min ]
box_y = [q1, q1, q3, q3, q1 ]
# make our median line vectors
med_x = [box_x_min, box_x_max]
# calculate 'notch' plot
else:
raise NotImplementedError
notch_max = med #+ 1.57*iq/np.sqrt(row)
notch_min = med #- 1.57*iq/np.sqrt(row)
if notch_max > q3:
notch_max = q3
if notch_min < q1:
notch_min = q1
# make our notched box vectors
box_x = [box_x_min, box_x_max, box_x_max, cap_x_max, box_x_max,
box_x_max, box_x_min, box_x_min, cap_x_min, box_x_min,
box_x_min ]
box_y = [q1, q1, notch_min, med, notch_max, q3, q3, notch_max,
med, notch_min, q1]
# make our median line vectors
med_x = [cap_x_min, cap_x_max]
med_y = [med, med]
doplot = plt.plot
whiskers.extend(doplot(wisk_x, [q1, wisk_lo], color=whiskerscolor, linestyle='--'))
whiskers.extend(doplot(wisk_x, [q3, wisk_hi], color=whiskerscolor, linestyle='--'))
caps.extend(doplot(cap_x, [wisk_hi, wisk_hi], color=capscolor, linestyle='-'))
caps.extend(doplot(cap_x, [wisk_lo, wisk_lo], color=capscolor, linestyle='-'))
boxes.extend(doplot(box_x, box_y, color=boxescolor, linestyle='-'))
medians.extend(doplot(med_x, med_y, color=medianscolor, linestyle='-'))
fliers.extend(doplot(flier_hi_x, flier_hi, sym,
flier_lo_x, flier_lo, sym))
# fix our axes/ticks up a little
newlimits = min(positions)-0.5, max(positions)+0.5
plt.gca().set_xlim(newlimits)
plt.gca().set_xticks(positions)
return dict(whiskers=whiskers, caps=caps, boxes=boxes,
medians=medians, fliers=fliers)
def ssi_scatter(timelock, iter = 100):
from myutils import bootstrap
from matplotlib.mlab import prctile, find
fig = plt.figure()
ax = fig.add_subplot(111)
storeSSI = {'PG':[],'FG':[]}
mem = {'PG response':'PG', 'response':'FG'}
units = timelock.units
# For each unit, compute the SSI for PG and FG
for unit in units:
data = timelock.get(unit)
# For now, I only want to look at hit-hit trials
select = (data['PG outcome']==consts['HIT']) & \
(data['outcome']==consts['HIT'])
trials = data[select]
inter = interval(unit,trials,'PG out','onset')
counts = inter[unit.id].map(len)
rates = counts/(inter['onset']-inter['PG out'])
for goal in mem.keys():
input = DataFrame({'rates':rates, 'stimulus':trials[goal]})
storeSSI[mem[goal]].append(bootstrap(input,ssi,iters=iter))
meanSSI = dict.fromkeys(storeSSI.keys())
intervSSI = dict.fromkeys(storeSSI.keys())
for key, ssis in storeSSI.iteritems():
# Calculate the means of the bootstrapped SSIs
meanSSI[key] = [ np.mean(unitSSI) for unitSSI in ssis ]
# Calculate the 95% confidence intervals of the boostrapped SSIs
intervSSI[key] = [ prctile(unitSSI,p=(2.5,97.5)) for unitSSI in ssis ]
# Now let's check for significance
sig = dict.fromkeys(meanSSI.keys())
def check_between(check, between):
is_it = (between[0] <= check) & (between[1] >= check)
return is_it
for key, iSSIs in intervSSI.iteritems():
sig[key] = np.array([ not check_between(0,issi) for issi in iSSIs ])
not_sig = [ not (pg | fg) for pg,fg in zip(sig['PG'],sig['FG']) ]
not_sig = np.array(not_sig)
sig_colors = {'PG':'r','FG':'b'}
xpnts = np.array(meanSSI['PG'])
ypnts = np.array(meanSSI['FG'])
xbars = np.abs(np.array(intervSSI['PG']).T - xpnts)
ybars = np.abs(np.array(intervSSI['FG']).T - ypnts)
# First, plot the not significant units
ax.errorbar(xpnts[not_sig],ypnts[not_sig],
yerr=ybars[:,not_sig],xerr=xbars[:,not_sig],
fmt='o', color = 'grey')
# Then plot things that are significant for PG and FG
for key in sig.iterkeys():
if sig[key].any():
ax.errorbar(xpnts[sig[key]],ypnts[sig[key]],
yerr=ybars[:,sig[key]],xerr=xbars[:,sig[key]],
fmt='o', color = sig_colors[key])
xs = ax.get_xlim()
ys = ax.get_ylim()
ax.plot(xs,[0,0],'-k')
ax.plot([0,0],ys,'-k')
ax.plot([-10,10],[-10,10],'--',color='grey')
ax.set_xlabel('PG SSI')
ax.set_ylabel('FG SSI')
ax.set_xlim(xs)
ax.set_ylim(ys)
ax.set_aspect('equal')
#fig.show()
return sig, not_sig
def make_tss_plot(bam_file, tss, prefix, chromsizes, read_len, bins=400, bp_edge=2000,
processes=8, greenleaf_norm=True):
'''
Take bootstraps, generate tss plots, and get a mean and
standard deviation on the plot. Produces 2 plots. One is the
aggregation plot alone, while the other also shows the signal
at each TSS ordered by strength.
'''
logging.info('Generating tss plot...')
tss_plot_file = '{0}_tss-enrich.png'.format(prefix)
tss_plot_data_file = '{0}_tss-enrich.txt'.format(prefix)
tss_plot_large_file = '{0}_large_tss-enrich.png'.format(prefix)
# Load the TSS file
tss = pybedtools.BedTool(tss)
tss_ext = tss.slop(b=bp_edge, g=chromsizes)
# Load the bam file
# Need to shift reads and just get ends, just load bed file?
bam = metaseq.genomic_signal(bam_file, 'bam')
bam_array = bam.array(tss_ext, bins=bins, shift_width=-read_len / 2, # Shift to center the read on the cut site
processes=processes, stranded=True)
# Actually first build an "ends" file
#get_ends = '''zcat {0} | awk -F '\t' 'BEGIN {{OFS="\t"}} {{if ($6 == "-") {{$2=$3-1; print}} else {{$3=$2+1; print}} }}' | gzip -c > {1}_ends.bed.gz'''.format(bed_file, prefix)
# print(get_ends)
# os.system(get_ends)
#bed_reads = metaseq.genomic_signal('{0}_ends.bed.gz'.format(prefix), 'bed')
# bam_array = bed_reads.array(tss_ext, bins=bins,
# processes=processes, stranded=True)
# Normalization (Greenleaf style): Find the avg height
# at the end bins and take fold change over that
if greenleaf_norm:
# Use enough bins to cover 100 bp on either end
num_edge_bins = int(100 / (2 * bp_edge / bins))
bin_means = bam_array.mean(axis=0)
avg_noise = (sum(bin_means[:num_edge_bins]) +
sum(bin_means[-num_edge_bins:] ))/(2*num_edge_bins)
bam_array /= avg_noise
else:
bam_array /= bam.mapped_read_count() / 1e6
# Generate a line plot
fig = plt.figure()
ax = fig.add_subplot(111)
x = np.linspace(-bp_edge, bp_edge, bins)
ax.plot(x, bam_array.mean(axis=0), color='r', label='Mean')
ax.axvline(0, linestyle=':', color='k')
# Note the middle high point (TSS)
tss_point_val = max(bam_array.mean(axis=0))
ax.set_xlabel('Distance from TSS (bp)')
ax.set_ylabel('Average read coverage (per million mapped reads)')
ax.legend(loc='best')
fig.savefig(tss_plot_file)
# Print a more complicated plot with lots of info
# write the plot data; numpy object
np.savetxt(tss_plot_data_file, bam_array.mean(axis=0), delimiter=",")
# Find a safe upper percentile - we can't use X if the Xth percentile is 0
upper_prct = 99
if mlab.prctile(bam_array.ravel(), upper_prct) == 0.0:
upper_prct = 100.0
plt.rcParams['font.size'] = 8
fig = metaseq.plotutils.imshow(bam_array,
x=x,
figsize=(5, 10),
vmin=5, vmax=upper_prct, percentile=True,
line_kwargs=dict(color='k', label='All'),
fill_kwargs=dict(color='k', alpha=0.3),
sort_by=bam_array.mean(axis=1))
# And save the file
fig.savefig(tss_plot_large_file)
return tss_plot_file, tss_plot_large_file, tss_point_val
def imshow(arr, x=None, ax=None, vmin=None, vmax=None, percentile=True,
strip=False, features=None, conf=0.95, sort_by=None,
line_kwargs=None, fill_kwargs=None, imshow_kwargs=None, figsize=(5, 12),
width_ratios=(4, 1), height_ratios=(4, 1),
subplot_params=dict(wspace=0.1, hspace=0.1),
subset_by=None, subset_order=None,):
"""
Do-it-all function to help with plotting heatmaps
Parameters
----------
arr : array-like
x : 1D array
X values to use. If None, use range(arr.shape[1])
ax : matplotlib.Axes
If not None, then only plot the array on the provided axes. This will
ignore any additional arguments provided that apply to figure-level
configuration or to the average line plot. For example, `figsize`,
`width_ratios`, `height_ratios`, `subplot_params`, `line_kwargs`, and
`fill_kwargs` will all be ignored.
vmin, vmax : float
percentile : bool
If True, then treat values for `vmin` and `vmax` as percentiles rather
than absolute values.
strip : bool
Include a strip plot alongside the array
features : pybedtools.BedTool or string filename
Features used to construct the array
conf : float
Confidence interval to use in line plot.
sort_by : array-like
Use the provided array to sort the array (e.g., an array of expression
values). This array will be argsorted to get the proper order.
line_kwargs, fill_kwargs : dict
Passed directly to `ci_plot`.
figsize : tuple
(Width, height) of the figure to create.
imshow_kwargs : dict
Passed directly to matplotlib.pyplot.imshow. By default, arguments
used are `origin='lower'`, `aspect="auto"` and a colormap from
colormap_adjust.smart_colormap generated using the provided `vmin` and
`vmax`.
width_ratios, height_ratios: tuple
These tuples are passed to the `new_shell` function. The default
values set up a 2x2 configuration of panels for heatmap, line plot,
colorbar axes, and optional strip plot. However modifying
`width_ratios` or `height_ratios` can be used to create more or fewer panels.
subplot_params : dict
Passed to Figure.subplots_adjust
subset_by : array
An array of any type (but usually int or str) that contains a class
label for each row in the heatmap array. For example, to subset by
expression, an array the values of "up", "down", or "unchanged" at each
of the positions could be provided.
Note that the heatmap array is first sorted by `sort_by` and then split
into groups according to `subset_by`, so each subset remains sorted by
`sort_by`.
subset_order : list-like
This provides the order in which the subsets are plotted. Since the
default imshow arguments contain `origin="lower"`, these will be
plotted in order starting at the bottom of the heatmap.
"""
if ax is None:
fig = new_shell(
figsize=figsize,
strip=strip,
subplot_params=subplot_params,
width_ratios=width_ratios,
height_ratios=height_ratios)
if x is None:
x = np.arange(arr.shape[1] + 1)
if percentile:
if vmin is None:
vmin = arr.min()
else:
vmin = mlab.prctile(arr.ravel(), vmin)
if vmax is None:
vmax = arr.max()
else:
vmax = mlab.prctile(arr.ravel(), vmax)
else:
if vmin is None:
vmin = arr.min()
if vmax is None:
vmax = arr.max()
cmap = colormap_adjust.smart_colormap(vmin, vmax)
_imshow_kwargs = dict(origin='lower', cmap=cmap, vmin=vmin, vmax=vmax,
aspect='auto')
if imshow_kwargs is not None:
_imshow_kwargs.update(imshow_kwargs)
# previously we did an argsort first; with subsetting we don't want to do
# that yet....
#if sort_by is not None:
# ind = np.argsort(sort_by)
#else:
# ind = np.arange(arr.shape[0])
if sort_by is None:
sort_by = np.arange(arr.shape[0])
if ax is None:
array_ax = fig.array_axes
else:
array_ax = ax
# If not provided, assume all in the same subset.
if subset_by is None:
subset_by = np.zeros(arr.shape[0])
# Ensure always array, since we're doing indexing tricks
if not isinstance(subset_by, np.ndarray):
subset_by = np.array(subset_by)
# If not provided, use sorted order
if subset_order is None:
subset_order = sorted(np.unique(subset_by))
inds = []
for cls in subset_order:
subset_ind = np.nonzero(subset_by == cls)[0]
subset_sort_by = sort_by[subset_ind]
subset_argsort_by = np.argsort(subset_sort_by)
inds.append(subset_ind[subset_argsort_by])
ind = np.concatenate(inds)
mappable = array_ax.imshow(
arr[ind, :],
extent=(x.min(), x.max(), 0, arr.shape[0]),
**_imshow_kwargs
)
if line_kwargs is None:
line_kwargs = {}
if fill_kwargs is None:
fill_kwargs = {}
if isinstance(line_kwargs, dict):
line_kwargs = [line_kwargs]
if isinstance(fill_kwargs, dict):
fill_kwargs = [fill_kwargs]
_line_kwargs = itertools.cycle(line_kwargs)
_fill_kwargs = itertools.cycle(fill_kwargs)
if ax is None:
plt.colorbar(mappable, fig.cax)
for subset_ind, label, _lkw, _fkw in zip(inds, subset_order, _line_kwargs, _fill_kwargs):
ci_plot(
x,
arr[subset_ind],
ax=fig.line_axes,
line_kwargs=_lkw,
fill_kwargs=_fkw,
)
return fig
else:
return ax.figure
#initializing function constant variables
N = Nmarkets * Nproducts
tolerance = 0.001
nogradient = 0
thetaspost = hlp.HMCMC(
lambda theta: hlp.computeGMMobjective(
theta, simshare, simoutshare, cdindex, weights, price, X, IV, vdraws,
Nproducts, N, tolerance, nogradient), theta0, B)
posteriormeanpost = np.mean(thetaspost[t - 1:, :], axis=0)
posteriormedianpost = np.median(thetaspost[t - 1:, :], axis=0)
thetasdemedianed = np.abs(thetaspost[t - 1:, :] - (np.ones(
(B - t + 1, 1)) @ posteriormedianpost[:, None].T))
criticalvaluesymmetricpost = matlab.prctile(thetasdemedianed,
100 * (1 - alpha))
posteriorquantilealpha2post = matlab.prctile(thetaspost[t - 1:, :],
100 * alpha / 2)
posteriorquantileoneminusalpha2post = matlab.prctile(thetaspost[t - 1:, :],
100 * (1 - alpha / 2))
### STANDARD ERRORS ###
betahat = np.zeros((1, dimX + 1))
for e in range(dimX + 1):
betanew = posteriormeanpost[e]
betahat[0, e] = betanew
betahat = betahat.conj().transpose()
theta2hat = np.zeros((1, dimX))
for c in range(dimX):
s.join(
"lista", 'lista_qtde_' +
datetime.now().strftime("%d%m%Y_%H_%M_%S") + '.csv'), 'w', 'utf-8')
for k, v in assuntos.items():
if 'valor' in v:
values.append(v['valor'])
line = "%s;%s\n" % (k, str(v['valor']))
print(line)
handle.write(line)
handle.close()
plt.figure()
#d = np.sort(np.random.randint(0, 1000, 1000)).cumsum()
d = sorted(values)
print(d)
# Percentile values
p = np.array([0.0, 25.0, 50.0, 75.0, 100.0])
perc = mlab.prctile(d, p=p)
plt.plot(d)
# Place red dots on the percentiles
plt.plot((len(d) - 1) * p / 100., perc, 'ro')
# Set tick locations and labels
plt.xticks((len(d) - 1) * p / 100., map(str, p))
plt.savefig(
s.join(
"figuras", 'resultado_perc_' +
datetime.now().strftime("%d%m%Y_%H_%M_%S") + '.png'))
def ssa(X, M=None, K=0):
r"""Performs Singular Spectrum Analysis on time series X with the method of
Vautard and Ghil, Phys. D. 1989.
Parameters
----------
X : 1D array
Vector of evenly spaced observations.
M : int
Window length. Default value is M = len(X) / 10
K : int
Number of EOFs used for reconstruction (AICC choice by default k=0).
if K = 0, corrected Akaike Information Criterion (AICC) is used
if K = 'mcssa', the Monte Carlo spectral significance estimation of
Allen & Smith (J Clim, 1996) is used.
Returns
-------
spec : array_like
Eigenvalue spectrum, in % variance.
eig_vec : array_like
Eigenvector matrix ("temporal EOFs").
PC : array_like
Matrix of principal components.
RC : array_like
Matrix of RCs (N*M, K) (only if K > 0).
RCp : array_like
Reconstructed time-series, involving only the modes retained, and
rescaled to original mean and variance.
Examples
--------
spec, eig_vec, PC, RC, RCp = ssa(X,[M, K])
Notes
-----
Orignal file hepta_ssa.m from Hepta Technologies, 2004 writing in MatlabTM.
last updated 03/14/2012 to include automated choice for K (AICC).
Julien Emile-Geay, Lamont Doherty Earth Observatory.
Dec 2004 last updated 03/14/2012
"""
X = np.atleast_1d(X)
if X.ndim > 1:
raise ValueError("Input vector `X` has more than 1 dimension.")
N = len(X)
# Center the series.
Xr, mu, sigma = standardize(X) # NOTE: Original calls standardize.m.
# Set default value for M.
if not M:
M = N // 10
if K == 'mcssa':
mcssa = True
MC = 1000
else:
mcssa = False
signif = np.arange(0, K) # FIXME: 0, K
Np = N - M + 1
gam, lags = xcorr(Xr, maxlags=M - 1, matlab_compat='unbiased')
# Fill in Covariance matrix. Take positive half of auto-correlation
# diagram, hence M to 2M - 1.
C = toeplitz(gam[M - 1:2 * M])
# Solve eigenvalue problem.
eig_vec, eig_val = eigd(C) # FIXME: Matlab eig_vec have reversed signs.
spec = eig_val / np.sum(eig_val)
# Determine significant eigenvalues.
if mcssa:
# NOTE: Got this at from: http://www.gps.caltech.edu/~tapio/arfit/
# But this is commented out in the original code.
#w, A, C, SBC, FPE, th = arfit(Xr, 1, 1) # fit AR(1) model.
# NOTE: The original code uses ar1.m.
# What is the difference between ar1.m and arfit.m?
a, var, _ = ar1(Xr)
s = np.sqrt(var)
noise = np.zeros(N, MC)
noise[0, :] = np.tile(Xr[0], np.r_[1, MC])
for jt in range(1, N):
noise[jt, :] = a * noise[jt - 1, :] + s * np.random.randn(1, MC)
noise, _, _ = standardize(noise)
Lambda_R = np.zeros_like(MC) # FIXME: Don't know the right shape yet.
for m in range(0, MC):
Gn, ln = xcorr(noise[:, m], M - 1, 'unbiased')
Cn = toeplitz(Gn[M: 2 * M - 1])
# Noise "eigenvalues".
tmp = np.dot(eig_vec, Cn)
Lambda_R[:, m] = np.diag(np.dot(tmp, eig_vec))
q95 = prctile(Lambda_R, 100 * 0.95) # FIXME
# Index of modes rising above the background.
signif = np.where(eig_val > q95)
print('MCSSA modes retained: %s' % signif)
fix, ax = plt.subplots()
ax.set_title('MCSSA')
v = np.arange[1, M + 1]
ligr = [0.7000, 0.7000, 0.7000]
lmin = Lambda_R.min(axis=1)
lmax = Lambda_R.max(axis=1)
ax.fill(v, lmin, lmax, ligr, ligr, 0, 0.3)
ax.plot(v, eig_val, 'kx', linewidth=2.0)
ax.plot(v, q95, 'r-', linewidth=2.0)
elif K == 0:
trunc = range(0, len(spec))
# The pca_truncation_criteria.m original call:
# [MDL, NE08, AIC, AICC] =
# pca_truncation_criteria(eig_val, 1, trunc, N, 1)
WK85, NE08 = pca_truncation_criteria(eig_val, 1, trunc, N, 1)
imin = (np.real(NE08['aicc'])).argmin()
K = trunc[imin]
print('AICC truncation choice, K = %s' % K)
signif = np.arange(0, K)
# Compute PCs.
decal = np.zeros((Np, M))
for t in range(0, N - M + 1):
decal[t, :] = Xr[t:M + t]
# The columns of this matrix are Ak(t), k=1 to M.
PC = np.dot(decal, eig_vec)
# Compute reconstructed timeseries if K > 0.
if signif:
RC = np.zeros((N, len(signif)))
# First M terms.
for t in range(0, M - 1):
Av = np.flipud(PC[0:t, signif])
eig_vec_red = eig_vec[0:t, signif]
RC[t, :] = 1.0 / t * np.sum(Av * eig_vec_red, axis=0)
# Middle of timeseries.
for t in range(M, Np + 1):
Av = np.flipud(PC[t - M + 1:t, signif])
eig_vec_red = eig_vec[0:M, signif]
RC[t, :] = 1 / M * np.sum(Av * eig_vec_red, axis=0)
# Last M terms.
for t in range(Np + 1, N + 1):
Av = np.flipud(PC[t - M + 1:Np, signif])
eig_vec_red = eig_vec[t - N + M:M, signif]
RC[t, :] = 1.0 / (N - t + 1) * np.sum(Av * eig_vec_red, axis=0)
# Sum and restore the mean and variance.
RCp = sigma * np.sum(RC, axis=1) + mu
else:
RC, RCp = None, None
return spec, eig_vec, PC, RC, RCp, signif
np.log10(faked_rdiffs))
std_over_shuffles = np.log10(faked_rdiffs).std(axis=0)
# floored 1-tailed p-value of actual std vs faked distr
std_n_more_extreme = np.sum(std_over_shuffles > real_std) + \
np.sum(~np.isfinite(std_over_shuffles))
mad_n_more_extreme = np.sum(mad_over_shuffles > real_mad)
assert np.sum(~np.isfinite(mad_over_shuffles)) == 0
std_pval = (std_n_more_extreme + 1) / float(faked_rdiffs.shape[1])
mad_pval = (mad_n_more_extreme + 1) / float(faked_rdiffs.shape[1])
# Text summary
sdump.append(region)
sdump.append("MAD, mean %0.3f, distr over shuffles: %s" % \
(mad_over_shuffles.mean(),
' ' .join(['%0.3f' % v for v in mlab.prctile(
mad_over_shuffles, (25, 50, 95, 97.5))])))
sdump.append("MAD, nP.E. actual=%0.3f, P.E. actual=%0.3f, pval=%0.6f" %
(real_npe_mad, real_mad, mad_pval))
sdump.append("STDEV, nanmean %0.3f, distr over shuffles: %s" % \
(np.nanmean(std_over_shuffles),
' '.join(['%0.3f' % v for v in mlab.prctile(
std_over_shuffles, (25, 50, 95, 97.5))])))
sdump.append("STDEV, nP.E. actual=%0.3f, P.E. actual=%0.3f, p=%0.6f" %
(real_npe_std, real_std, std_pval))
sdump.append('')
# Pretty
axa[0, 0].set_yticks((0, 1, 2, 3, 4))
axa[0, 1].set_yticks((0, 2, 4, 6))
axa[1, 0].set_yticks((0, 1, 2, 3, 4))
axa[1, 1].set_yticks((0, 2, 4, 6))
def sea(self, **kwargs):
"""Method called to perform superposed epoch analysis on data in object.
Uses object attributes obj.data, obj.times, obj.epochs, obj.delta,
obj.window, all of which must be available on instantiation.
Other Parameters
================
storedata : boolean
saves matrix of epoch windows as obj.datacube (default = False)
quartiles : list
calculates the quartiles as the upper and lower bounds (and is default);
ci : float
will find the bootstrapped confidence intervals of ci_quan at the ci percent level (default=95)
mad : float
will use +/- the median absolute deviation for the bounds;
ci_quan : string
can be set to 'median' (default) or 'mean'
Notes
=====
A basic plot can be raised with :meth:`plot`
"""
#check this hasn't already been done
#TODO: find out why doing two .sea() calls back-to-back fails 2nd time
if hasattr(self, 'semedian') or hasattr(self, 'semean'):
return None
#check defaults
defaults = {
'storedata': True,
'quartiles': True,
'ci': False,
'mad': False,
'ci_quan': 'median'
}
for default in defaults:
if default not in kwargs:
kwargs[default] = defaults[default]
#ensure all input is np array
delt = float(self.delta)
if isinstance(self.data, np.ndarray):
y = self.data
else:
y = np.asarray(self.data, dtype=float)
if kwargs['ci']:
kwargs['quartiles'], kwargs['mad'] = False, False
if kwargs['mad']:
kwargs['quartiles'], kwargs['ci'] = False, False
time, t_epoch = self._timeepoch(delt)
#build SEA matrix and perform analysis
wind = int(self.window)
m = int(2 * wind + 1)
n = len(t_epoch)
y_sea = np.zeros((n, m), dtype=float)
blankslice = np.zeros([m], dtype=float)
for i in range(n):
dif = np.abs(time - t_epoch[i])
j = np.where(dif == np.min(dif))
stpt = j[0][0] - wind
enpt = j[0][0] + wind + 1
sea_slice = blankslice.copy()
if stpt < 0: #fix for bad epochs not correctly moved to badepochs attr #TODO: make badepochs robust or do all checking here
sea_slice[0:abs(stpt)] = np.NaN
sea_slice[abs(stpt):] = y[0:enpt]
elif enpt >= len(y):
tmpslice = y[stpt:]
sea_slice[:len(tmpslice)] = tmpslice
sea_slice[len(tmpslice):] = np.NaN
else:
sea_slice = y[stpt:enpt]
y_sea[i, 0:] = sea_slice
#find SEA mean, median and percentiles - exclude NaNs (or badval)
try:
badval = kwargs['badval']
except KeyError:
badval = np.nan
y_sea_m = ma.masked_where(np.isnan(y_sea), y_sea)
else:
y_sea_m = ma.masked_values(y_sea, badval)
self.semean = [np.mean(y_sea_m[:, i].compressed()) for i in range(m)]
self.semedian = [
np.median(y_sea_m[:, i].compressed()) for i in range(m)
]
self.semean, self.semedian = np.array(self.semean), np.array(
self.semedian)
self.bound_low = np.zeros((m, 1))
self.bound_high = np.zeros((m, 1))
if kwargs['quartiles']:
from matplotlib.mlab import prctile
for i in range(m):
dum = np.sort(y_sea_m[:, i].compressed())
qul = prctile(dum, p=(25, 75))
self.bound_low[i], self.bound_high[i] = qul[0], qul[1]
self.bound_type = 'quartiles'
elif kwargs['ci']: #bootstrapped confidence intervals (95%)
funcdict = {'mean': np.mean, 'median': np.median}
try:
if isinstance(kwargs['ci'], bool):
raise ValueError #fall through to default case
else:
ci_level = float(kwargs['ci'])
except ValueError:
ci_level = 95
from spacepy.poppy import boots_ci
if hasattr(kwargs['ci_quan'], "__call__"): #ci_quan is a function
ci_func = kwargs['ci_quan']
else:
ci_func = funcdict[kwargs['ci_quan']]
for i in range(m):
dum = np.sort(y_sea_m[:, i].compressed())
self.bound_low[i], self.bound_high[i] = \
boots_ci(dum, 800, ci_level, ci_func)
self.bound_type = 'ci'
elif kwargs['mad']: #median absolute deviation
for i in range(m):
dum = np.sort(y_sea_m[:, i].compressed())
spread_mad = tb.medAbsDev(dum)
self.bound_low[i] = self.semedian[i] - spread_mad
self.bound_high[i] = self.semedian[i] + spread_mad
self.bound_type = 'mad'
self.x = np.linspace(-1.*self.window*self.delta, self.window*self.delta, \
len(self.semedian))
if kwargs['storedata']:
self.datacube = y_sea_m
if self.verbose:
print('sea(): datacube added as new attribute')
if self.verbose:
print('Superposed epoch analysis complete')