def fit(self, output='list', **kw):
''' needs scipy.optimize imported as opt
'''
# selection of optimizaiton method
method, options, tol = self.optimization_options(kw)
# bounds = ((0, None), (0, None), (0, None), (0, 1e3))
result = opt.minimize(self.chisqfc, self.params, method=method, options=options, tol=tol) # bounds=bounds
self.params = np.array(result.x)
self.results.append(result.x) # appending results to self.results
if output == 'list': # fit will return [params_BM[:], params_d[:]] as a unified list
return self.params
elif output == 'split': # fit will return [[params_BM], [params_d]]
return parameterSplit(self.params)
elif output == 'combo':
split = parameterSplit(self.params)
return self.params, split[0], split[1]
def chisqfc(self, parameters):
''' Calculates the sum of the chisq of all the bands'''
# preparing inputs
params_BM, params_d = parameterSplit(parameters)
chisq_sum = 0
chisq_log = []
for index, freq in enumerate(self.freqs): # iterates through the frequency bands
# finding the residual
residual = self.residual(index, freq, params_BM, params_d)
# finding the error
d_data = self.measured.d_fitdata[-1][index]
try: # checking if d_data exists
residual/d_data
except TypeError: # else assuming 2% error
d_data = [x/50 for x in self.measured.fitdata[-1][index]]
# finding the chi-square
chisq = np.sum((residual/d_data)**2)
# recording the chi-square
chisq_log.append(chisq)
chisq_sum += chisq
self.chisq.append(np.array(chisq_log))
return chisq_sum
def plotCorrelation(MCData, paramsList, name='MC Parameter Correlations.png'):
# Finding plot properties
numParams = len(MCData.params) # number of parameters
numRows, numCols = numParams, numParams # number of columns and rows
plots_per_row = [numParams-x for x in range(numParams)] # number of plots per row
numIterations = len(MCData.results[:, 0]) # number of data points
if numIterations < 100: # packaging data points into bins
binNumber = int(numIterations/10)
else:
binNumber = 100
fig = plt.figure()
gs = GridSpec(numRows, numCols)
for row, numPlots in enumerate(plots_per_row): # rows as indices and number of plots for the columns
plotStart, plotEnd = row, row+1 # determines the column start and stop point
for col in range(row, numPlots+row): # iterates through the parameters
# making axes
ax = plt.subplot(gs[row, plotStart:plotEnd]) # placing axes
plotStart, plotEnd = plotEnd, plotEnd+1 # moving next axes over a column
# finding x and y data
xdata = MCData.results[:, row] # selecting xdata
ydata = MCData.results[:, col] # selecting ydata
# correlation plot
if row == col:
histoData = np.histogram(MCData.results[:, row], binNumber) # *** when MCData packaged by params( not 1 2-D-list) will need different index ***
histogram = ax.bar(histoData[1][0:-1], histoData[0], histoData[1][1]-histoData[1][0], label='Hist')
mean = ax.axvline(x=MCData.params[row], c='g', linestyle='dashed', label='Mean')
# get right indices
splitparams = parameterSplit(MCData.params)
index1, index2 = 0, 0
for i in range(len(splitparams)):
for j in splitparams[i]:
if row == (index1+index2):
parameter = ax.axvline(x=paramsList[index1][index2], c='r', linestyle='dashed', label='Parameter')
plt.title('Parameter {},{}'.format(index1, index2)) # *** FIX THIS ***
break
index2 += 1
index2 = 0
index1 += 1
else:
correlplot, = ax.plot(xdata, ydata, 'bo', label='correlation')
# Best Fit Plot
plotList = np.arange(min(xdata), max(xdata), 0.1)
fitCoeffs = polyfit(xdata, ydata, 1)
bestfit, = ax.plot(plotList, [fitCoeffs[0] + fitCoeffs[1]*x for x in plotList], 'r-', label='bestfit')
# mean plots
param1mean = ax.axvline(x=np.mean(xdata), c='k', linestyle='dashed', linewidth=1, label='param{} mean: {:.2f}'.format(row, np.mean(xdata)))
param2mean = ax.axhline(y=np.mean(ydata), c='k', linestyle='dashed', linewidth=1, label='param{} mean: {:.2f}'.format(col+1, np.mean(ydata)))
plt.title('parameter {},{}'.format(col, row))
# plot properties
# plt.xlabel('parameter {}'.format(row+1))
# plt.ylabel('parameter {}'.format(col+2))
# handles, labels = ax.get_legend_handles_labels()
# ax.legend(handles, labels, loc='upper right')
ax.set_xticklabels([]) # turns off x-tick labels
ax.set_yticklabels([]) # turns off y-tick labels
# figure properties
plt.suptitle("Correlation Plots")
plt.savefig(name)
plt.close(fig)
return blackbody_Nu
def blackbody_convert(nu, List, T):
""" A Plancks Law frequency function to convert the Black Body equation to the right units
nu, const = [h, c, k, TVac, TDust]
TDust=19.6, h=6.62606957*(10**-34), c=299792458, k=1.3806488*(10**-23), TVac = 2.7
It is the derivative of blackbody_nu
"""
h, c, k, TVac, TDust = List[:]
numerator = 2 * (h ** 2) * (nu ** 4) * np.exp((h * nu) / (k * TVac))
denominator = k * (TVac ** 2) * (c ** 2) * ((np.exp((h * nu) / (k * TVac)) - 1) ** 2)
return numerator / denominator
##########################################################################################
# Total Signal
def totalSignal((lList, const, frequency, BMode_data), parameters): # Change name of inputs
""" Creates the total theoretical signal
"""
# creating parameter inputs
params_BM, params_d = parameterSplit(parameters)
# getting functions
BMode = BModeSignal(BMode_data, params_BM)
Dust = dustSignal((lList, const, frequency), params_d)
# adding into one signal
signal = np.array(BMode + Dust)
return signal
return blackbody_Nu
def blackbody_convert(nu, List, T):
''' A Plancks Law frequency function to convert the Black Body equation to the right units
nu, const = [h, c, k, TVac, TDust]
TDust=19.6, h=6.62606957*(10**-34), c=299792458, k=1.3806488*(10**-23), TVac = 2.7
'''
h, c, k, TVac, TDust = List[:]
return 2 * (h**2) * (nu**4) * np.exp(
(h * nu) / (k * TVac)) / (k * (TVac**2) * (c**2) * ((np.exp(
(h * nu) / (k * TVac)) - 1)**2))
##########################################################################################
# Total Signal
def totalSignal((lList, const, frequency, BMode_data), parameters
): # Change name of inputs
''' Creates the total theoretical signal
'''
# creating parameter inputs
params_BM, params_d = parameterSplit(parameters)
# getting functions
BMode = BModeSignal(BMode_data, params_BM)
dust = dustSignal((lList, const, frequency), params_d)
# adding into one signal
signal = dust + BMode
return np.array(signal)
def plotCorrelation(MCData, paramsList, name='MC Parameter Correlations.png'):
# Finding plot properties
numParams = len(MCData.params) # number of parameters
numRows, numCols = numParams, numParams # number of columns and rows
plots_per_row = [numParams - x
for x in range(numParams)] # number of plots per row
numIterations = len(MCData.params_log[:, 0]) # number of data points
if numIterations < 100: # packaging data points into bins
binNumber = int(numIterations / 10)
else:
binNumber = 100
fig = plt.figure()
gs = GridSpec(numRows, numCols)
for row, numPlots in enumerate(
plots_per_row
): # rows as indices and number of plots for the columns
plotStart, plotEnd = row, row + 1 # determines the column start and stop point
for col in range(row,
numPlots + row): # iterates through the parameters
# making axes
ax = plt.subplot(gs[row, plotStart:plotEnd]) # placing axes
plotStart, plotEnd = plotEnd, plotEnd + 1 # moving next axes over a column
# finding x and y data
xdata = MCData.params_log[:, row] # selecting xdata
ydata = MCData.params_log[:, col] # selecting ydata
if row == col: # histogram plot
histoData = np.histogram(MCData.params_log[:, row], binNumber)
histogram = ax.bar(histoData[1][0:-1],
histoData[0],
histoData[1][1] - histoData[1][0],
label='Hist')
mean = ax.axvline(x=MCData.params[row],
c='g',
linestyle='dashed',
label='Mean')
# get right indices
splitparams = parameterSplit(MCData.params)
index1, index2 = 0, 0
for i in range(len(splitparams)):
for j in splitparams[i]:
if row == (index1 + index2):
parameter = ax.axvline(
x=paramsList[index1][index2],
c='r',
linestyle='dashed',
label='Parameter')
plt.title('Parameter {},{}'.format(index1, index2),
fontsize=10) # *** FIX THIS ***
break
index2 += 1
index2 = 0
index1 += 1
else: # correlation plot
correlplot, = ax.plot(xdata, ydata, 'bo', label='correlation')
# Best Fit Plot
# plotList = np.arange(min(xdata), max(xdata), 0.1)
# fitCoeffs = polyfit(xdata, ydata, 1)
# bestfit, = ax.plot(plotList, [fitCoeffs[0] + fitCoeffs[1]*x for x in plotList], 'r-', label='bestfit')
# mean plots
param1mean = ax.axvline(x=np.mean(xdata),
c='k',
linestyle='dashed',
linewidth=1,
label='param{} mean: {:.2f}'.format(
row, np.mean(xdata)))
param2mean = ax.axhline(y=np.mean(ydata),
c='k',
linestyle='dashed',
linewidth=1,
label='param{} mean: {:.2f}'.format(
col + 1, np.mean(ydata)))
plt.title('parameter {},{}'.format(col, row), fontsize=10)
# plot properties
# plt.xlabel('parameter {}'.format(row+1))
# plt.ylabel('parameter {}'.format(col+2))
# handles, labels = ax.get_legend_handles_labels()
# ax.legend(handles, labels, loc='upper right')
# ax.set_xticklabels([]) # turns off x-tick labels
plt.setp(ax.get_xticklabels()[::2], visible=False)
ax.tick_params(axis='x', labelsize=8)
plt.setp(ax.get_yticklabels()[::2], visible=False)
ax.tick_params(axis='y', labelsize=8)
# figure properties
plt.suptitle("Correlation Plots")
plt.savefig(name)
plt.close(fig)
