def gaussianfit(self, num_zeros, params, limitedmin, limitedmax, minpars,
maxpars): #wrapper for the fitting algorithm
ajuste = multigaussfit(self.velocity,
self.intensity,
ngauss=num_zeros / 2,
params=params,
limitedmin=limitedmin,
limitedmax=limitedmax,
minpars=minpars,
maxpars=maxpars,
quiet=True,
shh=True)
fit_params = ajuste[0][:]
return fit_params
def fit_gaussians(x, y, guesses=None, ncomps=1):
""" Fits multiple Gaussians to y(x).
Parameters
----------
x, y : array-like
Data.
guesses : array-like, optional
Initial guesses for fitting: [amp, x0, width] for each componenet. If
ncomps > len(guesses) / 3 then the ncomps will be lowered to
len(guesses) / 3.
ncomps : int
Number of Gaussians to fit
Returns
-------
model : array-like
model_comps : list
Model components
model_params : list
list of model_params
"""
import numpy as np
from gaussfitter import multigaussfit
def gauss(x, amp, x0, width):
return amp * np.exp(-(x-x0)**2 / (2 * width**2))
fit = multigaussfit(x, y, ngauss=int(ncomps), params=np.asarray(guesses))
model = fit[1]
model_params, model_comps = [], []
for i in xrange(len(fit[0]) / 3):
model_param = fit[0][3*i:3*i+3]
model_params.append(model_param)
model_comps.append(gauss(x, *model_param))
return model, model_comps, model_params
def fit_gaussians(x, y, guesses=None, ncomps=1):
""" Fits multiple Gaussians to y(x).
Parameters
----------
x, y : array-like
Data.
guesses : array-like, optional
Initial guesses for fitting: [amp, x0, width] for each componenet. If
ncomps > len(guesses) / 3 then the ncomps will be lowered to
len(guesses) / 3.
ncomps : int
Number of Gaussians to fit
Returns
-------
model : array-like
model_comps : list
Model components
model_params : list
list of model_params
"""
import numpy as np
from gaussfitter import multigaussfit
def gauss(x, amp, x0, width):
return amp * np.exp(-(x - x0)**2 / (2 * width**2))
fit = multigaussfit(x, y, ngauss=int(ncomps), params=np.asarray(guesses))
model = fit[1]
model_params, model_comps = [], []
for i in xrange(len(fit[0]) / 3):
model_param = fit[0][3 * i:3 * i + 3]
model_params.append(model_param)
model_comps.append(gauss(x, *model_param))
return model, model_comps, model_params
def multifit(self):
self.ngauss = len(self.guesses)/3
self.setfitspec()
if self.fitkwargs.has_key('negamp'): self.fitkwargs.pop('negamp')
mpp,model,mpperr,chi2 = gaussfitter.multigaussfit(
self.specplotter.vind[self.gx1:self.gx2],
self.spectofit[self.gx1:self.gx2],
err=self.errspec[self.gx1:self.gx2],
ngauss=self.ngauss,
params=self.guesses,
**self.fitkwargs)
self.chi2 = chi2
self.dof = self.gx2-self.gx1-self.ngauss*3
self.model = model
self.modelpars = mpp.tolist()
self.modelerrs = mpperr.tolist()
self.modelplot = self.specplotter.axis.plot(
self.specplotter.vind[self.gx1:self.gx2],
self.model+self.specplotter.offset, color=self.fitcolor, linewidth=0.5)
self.residuals = self.spectofit[self.gx1:self.gx2] - self.model
if self.autoannotate:
self.annotate()
def multifit(self):
self.ngauss = len(self.guesses)/3
self.setfitspec()
if self.fitkwargs.has_key('negamp'): self.fitkwargs.pop('negamp')
mpp,model,mpperr,chi2 = gaussfitter.multigaussfit(
self.specplotter.vind[self.gx1:self.gx2],
self.spectofit[self.gx1:self.gx2],
err=self.errspec[self.gx1:self.gx2],
ngauss=self.ngauss,
params=self.guesses,
**self.fitkwargs)
self.chi2 = chi2
self.dof = self.gx2-self.gx1-self.ngauss*3
self.model = model
self.modelpars = mpp.tolist()
self.modelerrs = mpperr.tolist()
self.modelplot = self.specplotter.axis.plot(
self.specplotter.vind[self.gx1:self.gx2],
self.model+self.specplotter.offset, color=self.fitcolor, linewidth=0.5)
self.residuals = self.spectofit[self.gx1:self.gx2] - self.model
if self.autoannotate:
self.annotate()
def gaussparty(gausspars, nspec, CCFvaluelist, CCFxaxis, error_array, ngauss=2,
fixed=[False, False, False],
limitedmin=[False, False, True],
limitedmax=[False, False, False],
minpars=[0, 0, 0], maxpars=[0, 0, 0]):
'''
Fit 2 gaussians to some data.
This function is old and doesn't work very well. Use gaussbetter instead.
Parameters
----------
gausspars : `str`
Filename containing initial Gaussian parameter guesses
nspec : `int`
Number of spectra (visits)
CCFvaluelist : `list`
2D list with CCF values, one per visit
CCFxaxis : `list`
2D list with CCF RV x-axis values corresponding to CCFvaluelist, one per visit
error_array: `list`
2D list with errors corresponding to CCFvaluelist
fixed : `list` of bools
Is parameter fixed? (amplitude, position, width)
limitedmin/minpars : `list`
Set lower limits on each parameter (default: width > 0)
limitedmax/maxpars : `list`
Set upper limits on each parameter
minpars : `list` (amplitude, position, width)
maxpars : `list` (amplitude, position, width)
Returns
-------
gaussfitlist : `list`
2D list of results from calling multigaussfit
'''
assert len(CCFvaluelist) == nspec
assert len(CCFxaxis) == nspec
assert len(error_array) == nspec
param = []
with open(gausspars) as f1:
for line in f1:
if line[0] != '#':
param.append(line.rstrip())
assert len(param) == nspec
gaussfitlist = []
print(' ')
print('Gaussian fit results: peak amplitude, width, rvraw, rvraw_err')
print ('-------------------------------------------------------------')
for i in range(0, nspec):
if '#' in param[i]:
commentbegin = param[i].find('#')
partest = param[i][0:commentbegin].split()
else:
partest = param[i].split()
partest = [float(item) for item in partest]
gaussfit = gf.multigaussfit(CCFxaxis[i], CCFvaluelist[i], ngauss=ngauss,
params=partest, err=error_array[i], fixed=fixed,
limitedmin=limitedmin, limitedmax=limitedmax,
minpars=minpars, maxpars=maxpars,
quiet=True, shh=True, veryverbose=False)
gaussfitlist.append(gaussfit)
print('{0:.3f} {1:.2f} {2:.4f} {3:.4f} \t {4:.3f} {5:.2f} {6:.4f} {7:.4f}'.format(
gaussfit[0][0], gaussfit[0][2], gaussfit[0][1], gaussfit[2][1],
gaussfit[0][3], gaussfit[0][5], gaussfit[0][4], gaussfit[2][4]))
return gaussfitlist
def gaussparty(gausspars, nspec, filenamelist, bfsmoothlist, bf_ind, amplimits, threshold, widlimits):
'''
Fits 2 or 3 gaussians to some data
'''
param = []
with open(gausspars) as f1:
for line in f1:
if line[0] != '#':
param.append( line.rstrip() )
#param = np.loadtxt(gausspars, comments='#')
bffitlist = []
bffitlist.append(0)
gauss1 = [[] for i in range(nspec)]
gauss2 = [[] for i in range(nspec)]
gauss3 = [[] for i in range(nspec)]
gauss1[0] = [0,0]
gauss2[0] = [0,0]
gauss3[0] = [0,0]
error_array = np.ones(len(bfsmoothlist[0]))*0.01 # dummy array with 0.01 error values
print(' ')
print('Gaussian fit results: peak amplitude, width, rvraw, rvraw_err')
print ('-------------------------------------------------------------')
for i in range(1, nspec):
# check to see if we are fitting a third gaussian, i.e., one near zero
# don't print out the result of this fit, but do return it for plotting
# handle comments in gausspars file without exploding
if '#' in param[i]:
commentbegin = param[i].find('#')
partest = param[i][0:commentbegin].split()
else:
partest = param[i].split()
if len(partest) == 6: ngauss = 2
elif len(partest) == 9: ngauss = 3
else: print('something is wrong with your gausspars file!')
# min and max pars for peak 1: amp, rv, width
#minpars=[0.8, float(partest[1])-threshold, 0]
#maxpars=[1.0, float(partest[1])+threshold, 7]
# min and max pars for peak 2: amp, rv, width
#minpars.extend([0.05, float(partest[4])-threshold, 0])
#maxpars.extend([0.20, float(partest[4])+threshold, 40])
minpars = [amplimits[0], float(partest[1])-threshold, widlimits[0]]
maxpars = [amplimits[1], float(partest[1])+threshold, widlimits[1]]
minpars.extend([amplimits[2], float(partest[4])-threshold, widlimits[2]])
maxpars.extend([amplimits[3], float(partest[4])+threshold, widlimits[3]])
if ngauss == 2:
bffit = gf.multigaussfit(bf_ind, bfsmoothlist[i], ngauss=ngauss,
params=partest, err=error_array,
limitedmin=[True,True,True], limitedmax=[True,True,True],
minpars=minpars, maxpars=maxpars, quiet=True, shh=True)
elif ngauss == 3:
# min and max pars for peak 3: amp, rv, width (hardwired)
minpars.extend([0.05, float(partest[7])-threshold, 0])
maxpars.extend([1.0, float(partest[7])+threshold, 30])
bffit = gf.multigaussfit(bf_ind, bfsmoothlist[i], ngauss=ngauss,
params=partest, err=error_array,
limitedmin=[True,True,True], limitedmax=[True,True,True],
minpars=minpars, maxpars=maxpars, quiet=True, shh=True)
newbffit = [[] for x in range(len(bffit))]
# Sometimes bffit[2] is None, or contains None. Set it to zeros instead.
try:
if not any(bffit[2]): # this will fail if bffit[2] = None
newbffit[0] = bffit[0]
newbffit[1] = bffit[1]
newbffit[2] = [0, 0, 0, 0, 0]
else:
newbffit = bffit
except:
print('WARNING - gaussfit is acting up, fit failed for the next row, adjust gausspars file:')
if not bffit[2]: # this catches the case where bffit[2] = None
newbffit[0] = bffit[0]
newbffit[1] = bffit[1]
newbffit[2] = [0, 0, 0, 0, 0]
else:
newbffit = bffit
bffitlist.append(newbffit)
# NOTE: to get the gaussian fit corresponding to bfsmoothlist[i], use bffitlist[i][1].
# RV1 for observation i is bffitlist[i][0][1] +/- bffitlist[i][2][1].
# RV2 for observation i is bffitlist[i][0][4] +/- bffitlist[i][2][4].
# (note: need to check if bffit[2] == None before calling bffit[2][1] or bffit[2][4])
print('{0:s} {1:.3f} {2:.2f} {3:.4f} {4:.4f} \t {5:.3f} {6:.2f} {7:.4f} {8:.4f}'.format(
filenamelist[i][-20:], newbffit[0][0], newbffit[0][2], newbffit[0][1], newbffit[2][1],
newbffit[0][3], newbffit[0][5], newbffit[0][4], newbffit[2][4]))
print(' ')
print('You MUST manually guesstimate the location of each Gaussian\'s peak in %s!' % gausspars)
print('Until you do, the above values will be WRONG and the plot will look TERRIBLE.')
print(' ')
return bffitlist
def main(argv):
inputfile = ''
outputfile = ''
try:
opts, args = getopt.getopt(argv,"hi:j:o:",["ifile=","ifile2","ofile="])
except getopt.GetoptError:
print 'test.py -i <inputfile> -j <inputfile2> -o <outputfile>'
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print 'test.py -i <inputfile> -j <inputfile2> -o <outputfile>'
sys.exit()
elif opt in ("-i", "--ifile"):
inputfile = arg
elif opt in ("-j", "--ifile2"):
inputfile2 = arg
elif opt in ("-o", "--ofile"):
outputfile = arg
# data
f = open (inputfile, "rt")
#print inputfile
l = f.readline()
time = []
counts = []
is_traj = 0
while l:
va = l.split()
if (len(va)==5 and va[0]=="record#"):
is_traj = 1
l = f.readline()
va = l.split()
if is_traj == 1:
time.append(float(va[0]))
counts.append(float(va[1]))
l = f.readline()
f.close() #close the data file
# data thresh
f2 = open (inputfile2, "rt")
l2 = f2.readline()
th_u = []
th_d = []
f3 = open('../tmp/gaussFitTEX.txt','w')
while l2:
va = l2.split()
th_u.append(float(va[0]))
th_d.append(float(va[1]))
l2 = f2.readline()
## calculate histogram
## now determine nice limits by hand:
binwidth = 1
xmax = np.max(np.fabs(time))
ymax = np.max(np.fabs(counts))
bins = np.arange(-0.5, ymax + binwidth-0.5, binwidth)
hist_y, hist_x = np.histogram(counts, bins=bins)#, normed = True
hist_x = bins + 0.5
#find init pars
first_x = hist_x[0:int(th_u[0])]
first_y = hist_y[0:int(th_u[0])]
second_x = hist_x[int(th_u[0]):]
second_y = hist_y[int(th_u[0]):]
#print th_u[0]
#print first_x[:]
#print second_x[0]
#print max(first_y)
#print max(second_y)
##print hist_x.index(20.0)
#print np.where(first_y==max(first_y))
#print np.where(hist_y==max(second_y))
a_f = max(first_y) #amplitude for first max
o_f = np.where(first_y==max(first_y))[0][0] # offset for first max
w_f = mt.sqrt(o_f) #width for first max
#print array[tmp_where[0][0]]
a_s = max(second_y)#amplitude for second max
o_s = np.where(hist_y==max(second_y))[0][-1]# offset for second max
w_s = mt.sqrt(o_s)#width for first max
#print a_f
#print o_f
#print w_f
#print a_s
#print o_s
#print w_s
# parameters: ampl, offset, width
limits = [0.,False, False]
AIC1 = []
AIC2 = []
print "2 Gauss fit"
ngauss_now = 2
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss = ngauss_now, params=[a_f,o_f,w_f, a_s,o_s,w_s])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss = ngauss_now, params=[a_f,o_f,w_f, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
#-----------------------------------------------------------------
#!!!!!!!!!!!!!!!!!! I can also add errors!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
#-----------------------------------------------------------------
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&&&&&&& \\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&&&&&&& \\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&&&&&&& \\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,1)
plot_fits(hist_x, hist_y, n_gauss, pars)
#f3.write(pars)
#print "pars"
#print pars
#print "n_gauss"
#print n_gauss
#print "errors"
#print errors
#print "chii"
#print chii/len(hist_y)
print "3 Gauss fit"
ngauss_now = 3
o_m = int((o_f+o_s)/2)
a_m = hist_y[o_m]
w_m = mt.sqrt(o_m)
#print o_m
#print a_m
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss = ngauss_now, params=[a_f,o_f,w_f, a_m,o_m,w_m, a_s,o_s,w_s])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss = ngauss_now, params=[a_f,o_f,w_f, a_m,o_m,w_m, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&&&&&& \\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&&&&&& \\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&&&&&& \\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,2)
plot_fits(hist_x, hist_y, n_gauss, pars)
#print chii/len(counts)
#print "pars"
#print pars
#print "n_gauss"
#print n_gauss
#print "errors"
#print errors
#print "chii"
#print chii/len(hist_y)
print "4 Gauss fit"
ngauss_now = 4
o_1 = int((o_f+o_s)/4)
a_1 = hist_y[o_1]
w_1 = mt.sqrt(o_1)
o_2 = int(3*(o_f+o_s)/4)
a_2 = hist_y[o_2]
w_2 = mt.sqrt(o_2)
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y, ngauss = ngauss_now, params=[a_f,o_f,w_f, a_1,o_1,w_1, a_2,o_2,w_2, a_s,o_s,w_s])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y, ngauss = ngauss_now, params=[a_f,o_f,w_f, a_1,o_1,w_1, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&&&&& \\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&&&&& \\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&&&&& \\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,3)
plot_fits(hist_x, hist_y, n_gauss, pars)
#print chii/len(counts)
print "5 Gauss fit"
ngauss_now = 5
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&' + "%.3f" %(pars[12]) + '&&&& \\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&' + "%.3f" %(pars[3]) + '&&&& \\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&' + "%.3f" %(pars[14]) + '&&&& \\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&&\\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,4)
plot_fits(hist_x, hist_y, n_gauss, pars)
#print chii/len(counts)
print "6 Gauss fit"
ngauss_now = 6
o_0 = int(o_f*0.8)
a_0 = hist_y[o_0]
w_0 = mt.sqrt(o_0)
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&' + "%.3f" %(pars[12]) + '&' + "%.3f" %(pars[15]) + '&&&\\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[16]) + '&&&\\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&' + "%.3f" %(pars[14]) + '&' + "%.3f" %(pars[17]) + '&&&\\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,5)
plot_fits(hist_x, hist_y, n_gauss, pars)
print "7 Gauss fit"
ngauss_now = 7
o_n = int(o_s*1.2)
a_n = hist_y[o_n]
w_n = mt.sqrt(o_n)
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s, a_n,o_n,w_n])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&' + "%.3f" %(pars[12]) + '&' + "%.3f" %(pars[15]) + '&' + "%.3f" %(pars[18]) + '&&\\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[16]) + '&' + "%.3f" %(pars[19]) + '&&\\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&' + "%.3f" %(pars[14]) + '&' + "%.3f" %(pars[17]) + '&' + "%.3f" %(pars[20]) + '&&\\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,6)
plot_fits(hist_x, hist_y, n_gauss, pars)
print "8 Gauss fit"
ngauss_now = 8
o_5_1 = int((o_f+o_s)/5)
a_5_1 = hist_y[o_5_1]
w_5_1 = mt.sqrt(o_5_1)
o_5_2 = int(2*(o_f+o_s)/5)
a_5_2 = hist_y[o_5_2]
w_5_2 = mt.sqrt(o_5_2)
o_5_3 = int(3*(o_f+o_s)/5)
a_5_3 = hist_y[o_5_3]
w_5_3 = mt.sqrt(o_5_3)
o_5_4 = int(4*(o_f+o_s)/5)
a_5_4 = hist_y[o_5_4]
w_5_4 = mt.sqrt(o_5_4)
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_5_1,o_5_1,w_5_1, a_5_2,o_5_2,w_5_2, a_5_3,o_5_3,w_5_3, a_5_4,o_5_4,w_5_4, a_s,o_s,w_s, a_n,o_n,w_n])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&' + "%.3f" %(pars[12]) + '&' + "%.3f" %(pars[15]) + '&' + "%.3f" %(pars[18]) + '&' + "%.3f" %(pars[21]) + '& \\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[16]) + '&' + "%.3f" %(pars[19]) + '&' + "%.3f" %(pars[22]) + '& \\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&' + "%.3f" %(pars[14]) + '&' + "%.3f" %(pars[17]) + '&' + "%.3f" %(pars[20]) + '&' + "%.3f" %(pars[23]) + '& \\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,7)
plot_fits(hist_x, hist_y, n_gauss, pars)
print "9 Gauss fit"
ngauss_now = 9
o_s2 = int(o_s*1.1)
a_s2 = hist_y[o_s2]
w_s2 = mt.sqrt(o_s2)
(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_5_1,o_5_1,w_5_1, a_5_2,o_5_2,w_5_2, a_5_3,o_5_3,w_5_3, a_5_4,o_5_4,w_5_4, a_s,o_s,w_s, a_s2,o_s2,w_s2, a_n,o_n,w_n])
#(pars,n_gauss,errors, chii) = multigaussfit(hist_x[:-1],hist_y,ngauss=ngauss_now, params=[a_0,o_0,w_0, a_f,o_f,w_f, a_1,o_1,w_1, a_m,o_m,w_m, a_2,o_2,w_2, a_s,o_s,w_s], limitedmin=limits*ngauss_now)
chi_norm = chii/len(hist_y)
f3.write('amplitude & ' + "%.3f" %(pars[0]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[6]) + '&' + "%.3f" %(pars[9]) + '&' + "%.3f" %(pars[12]) + '&' + "%.3f" %(pars[15]) + '&' + "%.3f" %(pars[18]) + '&' + "%.3f" %(pars[21]) + '&' + "%.3f" %(pars[24]) + '\\\\\n')
f3.write('offset & ' + "%.3f" %(pars[1]) + '&' + "%.3f" %(pars[4]) + '&' + "%.3f" %(pars[7]) + '&' + "%.3f" %(pars[10]) + '&' + "%.3f" %(pars[3]) + '&' + "%.3f" %(pars[16]) + '&' + "%.3f" %(pars[19]) + '&' + "%.3f" %(pars[22]) + '&' + "%.3f" %(pars[25]) + '\\\\\n')
f3.write('width & ' + "%.3f" %(pars[2]) + '&' + "%.3f" %(pars[5]) + '&' + "%.3f" %(pars[8]) + '&' + "%.3f" %(pars[11]) + '&' + "%.3f" %(pars[14]) + '&' + "%.3f" %(pars[17]) + '&' + "%.3f" %(pars[20]) + '&' + "%.3f" %(pars[23]) + '&' + "%.3f" %(pars[26]) + '\\\\\n')
f3.write('chi&'+ "%.3f" %(chii) + '& chi norm&'+ "%.3f" %(chi_norm) + '&&&&& \\\\\\hline\n')
#AIC model selection
AIC1.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[0])
AIC2.append(calc_AIC(chii, ngauss_now*3, len(hist_y) )[1])
pl.subplot(2,4,8)
plot_fits(hist_x, hist_y, n_gauss, pars)
#print chii/len(counts)
minAIC1 = min(AIC1)
for l in range(len(AIC1)):
if (AIC1[l]==minAIC1):
AIC1[l] = '\\textbf{' + "%.2f" % (AIC1[l])+'}'
else:
AIC1[l] = "%.2f" % (AIC1[l])
AIC1_f = open('../GaussFit/PCH_modelSel/aic1Gauss_TEX.txt','a')
AIC1_f.write('$' + inputfile[12:] +'$&'+AIC1[0]+'&'+AIC1[1]+'&'+AIC1[2]+'&'+AIC1[3]+'&'+AIC1[4]+'&'+AIC1[5]+'&'+AIC1[6]+'&'+AIC1[7]+'\\\\\n')#
AIC1_f.close()
minAIC2 = min(AIC2)
for l in range(len(AIC2)):
if (AIC2[l]==minAIC2):
AIC2[l] = '\\textbf{' + "%.2f" % (AIC2[l])+'}'
else:
AIC2[l] = "%.2f" % (AIC2[l])
AIC2_f = open('../GaussFit/PCH_modelSel/aic2Gauss_TEX.txt','a')
AIC2_f.write('$' + inputfile[12:]+'$&'+AIC2[0]+'&'+AIC2[1]+'&'+AIC2[2]+'&'+AIC2[3]+'&'+AIC2[4]+'&'+AIC2[5]+'&'+AIC2[6]+'&'+AIC2[7]+'\\\\\n')#+'&'+AIC2[5]
AIC2_f.close()
pl.savefig(outputfile)
f3.closed
#mask_out.write(str(wavelength_angstroms[fit_cut])) plot = fig.add_subplot(2,2,2) plt.plot(wavelength_angstroms[first_cut], cont_subtracted_img[first_cut], color="blue", label="Data") redshift_array = np.array([0.0, 0.0, 0.0]) dispersion_array = np.array([0.0, 0.0, 0.0]) intensity_array = np.array([0.0, 0.0, 0.0]) peak_array = np.array([0.0, 0.0, 0.0]) for index, line in enumerate(wavelengths): p0 = [0.25, line, 7.5] if max(spectrum) > 10: output = gaussfitter.multigaussfit(wavelength_angstroms, cont_subtracted_img, ngauss=1, err=None, params=[0.25, line, 7.5], fixed=[False,False,False], limitedmin=[True,False,True], limitedmax=[False,False,True], minpars=[0,0,0], maxpars=[0,0,10.0], quiet=True, shh=True, veryverbose=False) coeff = output[0] gauss_line = gauss(wavelength_angstroms, *coeff) else: gauss_line = np.arange(len(spectrum)) coeff=[0,line,7.5] redshift_array[index] = (coeff[1]-emitted_wavelengths[index])/emitted_wavelengths[index] dispersion_array[index] = coeff[2] intensity_array[index] = coeff[0] peak_array[index] = coeff[1] #print(redshift_array) #stop()
def gaussparty(gausspars, nspec, filenamelist, bfsmoothlist, bf_ind, amplimits,
threshold, widlimits):
'''
Fits 2 or 3 gaussians to some data
'''
param = []
with open(gausspars) as f1:
for line in f1:
if line[0] != '#':
param.append(line.rstrip())
#param = np.loadtxt(gausspars, comments='#')
bffitlist = []
bffitlist.append(0)
gauss1 = [[] for i in range(nspec)]
gauss2 = [[] for i in range(nspec)]
gauss3 = [[] for i in range(nspec)]
gauss1[0] = [0, 0]
gauss2[0] = [0, 0]
gauss3[0] = [0, 0]
error_array = np.ones(len(
bfsmoothlist[0])) * 0.01 # dummy array with 0.01 error values
print(' ')
print('Gaussian fit results: peak amplitude, width, rvraw, rvraw_err')
print('-------------------------------------------------------------')
for i in range(1, nspec):
# check to see if we are fitting a third gaussian, i.e., one near zero
# don't print out the result of this fit, but do return it for plotting
# handle comments in gausspars file without exploding
if '#' in param[i]:
commentbegin = param[i].find('#')
partest = param[i][0:commentbegin].split()
else:
partest = param[i].split()
if len(partest) == 6: ngauss = 2
elif len(partest) == 9: ngauss = 3
else: print('something is wrong with your gausspars file!')
minpars = [amplimits[0], float(partest[1]) - threshold, widlimits[0]]
maxpars = [amplimits[1], float(partest[1]) + threshold, widlimits[1]]
minpars.extend(
[amplimits[2],
float(partest[4]) - threshold, widlimits[2]])
maxpars.extend(
[amplimits[3],
float(partest[4]) + threshold, widlimits[3]])
if ngauss == 2:
bffit = gf.multigaussfit(bf_ind,
bfsmoothlist[i],
ngauss=ngauss,
params=partest,
err=error_array,
limitedmin=[True, True, False],
limitedmax=[True, True, False],
minpars=minpars,
maxpars=maxpars,
quiet=True,
shh=True)
elif ngauss == 3:
# min and max pars for peak 3: amp, rv, width
# this will not work if len(amplimits) < 6 or len(widlimits) < 6
minpars.extend(
[amplimits[4],
float(partest[7]) - threshold, widlimits[4]])
maxpars.extend(
[amplimits[5],
float(partest[7]) + threshold, widlimits[5]])
bffit = gf.multigaussfit(bf_ind,
bfsmoothlist[i],
ngauss=ngauss,
params=partest,
err=error_array,
limitedmin=[True, True, True],
limitedmax=[True, True, True],
minpars=minpars,
maxpars=maxpars,
quiet=True,
shh=True)
newbffit = [[] for x in range(len(bffit))]
# Sometimes bffit[2] is None, or contains None. Set it to zeros instead.
try:
if not any(bffit[2]): # this will fail if bffit[2] = None
newbffit[0] = bffit[0]
newbffit[1] = bffit[1]
newbffit[2] = [0, 0, 0, 0, 0]
else:
newbffit = bffit
except:
print(
'WARNING - gaussfit is acting up, fit failed for the next row, adjust gausspars file:'
)
if not bffit[2]: # this catches the case where bffit[2] = None
newbffit[0] = bffit[0]
newbffit[1] = bffit[1]
newbffit[2] = [0, 0, 0, 0, 0]
else:
newbffit = bffit
bffitlist.append(newbffit)
# NOTE: to get the gaussian fit corresponding to bfsmoothlist[i], use bffitlist[i][1].
# RV1 for observation i is bffitlist[i][0][1] +/- bffitlist[i][2][1].
# RV2 for observation i is bffitlist[i][0][4] +/- bffitlist[i][2][4].
# (note: need to check if bffit[2] == None before calling bffit[2][1] or bffit[2][4])
if ngauss == 2:
print(
'{0:s} {1:.3f} {2:.2f} {3:.4f} {4:.4f} \t {5:.3f} {6:.2f} {7:.4f} {8:.4f}'
.format(filenamelist[i], newbffit[0][0], newbffit[0][2],
newbffit[0][1], newbffit[2][1], newbffit[0][3],
newbffit[0][5], newbffit[0][4], newbffit[2][4]))
elif ngauss == 3:
#\t {9:.3f} {10:.2f} {11:.4f} {12:.4f}'
print(
'{0:s} {1:.3f} {2:.2f} {3:.4f} {4:.4f} \t {5:.3f} {6:.2f} {7:.4f} {8:.4f}'
.format(filenamelist[i], newbffit[0][0], newbffit[0][2],
newbffit[0][1], newbffit[2][1], newbffit[0][3],
newbffit[0][5], newbffit[0][4], newbffit[2][4])
) #, newbffit[0][6], newbffit[0][8], newbffit[0][7], newbffit[2][7]))
print(' ')
print(
'You MUST manually guesstimate the location of each Gaussian\'s peak in %s!'
% gausspars)
print(
'Until you do, the above values will be WRONG and the plot will look TERRIBLE.'
)
print(' ')
return bffitlist
def gaussianfit(self,num_zeros,params,limitedmin,limitedmax,minpars,maxpars):#wrapper for the fitting algorithm ajuste=multigaussfit(self.velocity,self.intensity,ngauss=num_zeros/2,params=params,limitedmin=limitedmin,limitedmax=limitedmax,minpars=minpars,maxpars=maxpars,quiet=True,shh=True) fit_params=ajuste[0][:] return fit_params
def fit_110(wave, csubSI, csuberrSI, rms):
# print("-----------")
# print("FITTING 11.0")
# print("-----------")
is_SH_data = 0
wave_feat = wave
flux_feat = csubSI
fluxerr_feat = csuberrSI
if is_SH_data == 0:
my_ind = np.where((wave_feat > 10.6) & (wave_feat < 11.8))
w = wave_feat[my_ind]
f = flux_feat[my_ind]
e = fluxerr_feat[my_ind]
amp_guess = max(f[np.where(w < 11.1)])
params = [
10.989,
to_sigma(0.1538), amp_guess, 11.258,
to_sigma(0.236),
max(f)
],
limitedmin = [True, True, True, True, True, True],
limitedmax = [True, True, False, True, True, False],
minpars = [
10.97,
to_sigma(0.12), amp_guess / 20., 11.2,
to_sigma(0.226), 0.
],
maxpars = [11.05, to_sigma(0.19), 0., 11.3, to_sigma(0.246), 0.],
yfit = multigaussfit(w,
f,
err=e,
ngauss=2,
params=params,
limitedmin=limitedmin,
limitedmax=limitedmax,
minpars=minpars,
maxpars=maxpars,
quiet=True,
shh=True)
else:
my_ind = np.where((wave_feat > 10.6) & (wave_feat < 11.3))
w = wave_feat[my_ind]
f = flux_feat[my_ind]
e = fluxerr_feat[my_ind]
amp_guess = max(f[np.where(w < 11.1)])
params = [
10.989,
to_sigma(0.1538), amp_guess, 11.258,
to_sigma(0.236),
max(f)
],
limitedmin = [True, True, True, True, True, True],
limitedmax = [True, True, False, True, True, False],
minpars = [
10.97,
to_sigma(0.02), amp_guess / 20., 11.2,
to_sigma(0.16), 0.
],
maxpars = [11.05, to_sigma(0.19), 0., 11.3, to_sigma(0.246), 0.],
yfit = multigaussfit(w,
f,
err=e,
ngauss=2,
params=params,
limitedmin=limitedmin,
limitedmax=limitedmax,
minpars=minpars,
maxpars=maxpars,
quiet=True,
shh=True)
y1r = onedgaussian(w, 0, yfit[0][2], yfit[0][0], yfit[0][1])
# y2r = onedgaussian(w, 0, yfit[0][5], yfit[0][3], yfit[0][4])
# 11.0
small_gauss_area = sp.integrate.trapz(y1r, x=w)
position = yfit[0][0]
sigma = yfit[0][1]
amp = yfit[0][2]
# position_err = yfit[2][0]
sigma_err = yfit[2][1]
amp_err = yfit[2][2]
small_gauss_area_err = np.sqrt(
(amp_err / amp)**2 + (sigma_err / sigma)**2) * small_gauss_area
myrange = [position - (3. * sigma), position + (3. * sigma)]
N = np.where((wave_feat >= myrange[0]) & (wave_feat <= myrange[1]))[0]
dl = w[1] - w[0]
measured_flux_noise110 = (rms * np.sqrt(len(N)) * dl * 2)
# 11.2
# gauss_area = sp.integrate.trapz(y2r, x=w)
position = yfit[0][3]
sigma = yfit[0][4]
amp = yfit[0][5]
# position_err = yfit[2][3]
sigma_err = yfit[2][4]
amp_err = yfit[2][5]
# gauss_area_err = np.sqrt(
# (amp_err / amp)**2 + (sigma_err / sigma)**2) * gauss_area
myrange = [position - (3. * sigma), position + (3. * sigma)]
N = np.where((wave_feat >= myrange[0]) & (wave_feat <= myrange[1]))[0]
dl = w[1] - w[0]
measured_flux_noise112 = (rms * np.sqrt(len(N)) * dl * 2)
######################################
new_flux_feat = f - y1r
trap_flux_high = sp.integrate.trapz(new_flux_feat + e, x=w)
trap_flux_low = sp.integrate.trapz(new_flux_feat - e, x=w)
trap_flux = np.mean([trap_flux_high, trap_flux_low])
# trap_flux_std = 0.67 * np.std([trap_flux_high, trap_flux_low])
######################################
FINAL_112_FLUX = trap_flux # full_trap_flux - small_gauss_area
FINAL_112_FLUX_ERR = measured_flux_noise112 # + trap_flux_std
FINAL_110_FLUX = small_gauss_area
FINAL_110_FLUX_ERR = small_gauss_area_err + measured_flux_noise110
return FINAL_110_FLUX, FINAL_110_FLUX_ERR, \
FINAL_112_FLUX, FINAL_112_FLUX_ERR, myrange
