def test_fit(self):
# Test the method Class1.method1
# See http://docs.python.org/library/unittest.html#test-cases for further
# information
#print "run gaussian fit"
#print gaussian
fit, popt, perr = nonLinFit(gaussian, self.x_data, self.y_data, p_guess=self.p_guess, verbose=False)
diff = max(fit) - (self.p[1]+self.p[0])
self.assertTrue(abs(diff) < 0.0001)
for i in range(len(self.p)):
self.assertAlmostEqual(self.p[i],popt[i])
#print "run gauss + polynomial fit"
fitPoly, poptPoly, perr = nonLinFit(gaussPlusPoly, self.x_data, self.y_data_poly, p_guess=self.p_poly_guess, verbose=False)
for i in range(len(self.p)):
self.assertAlmostEqual(self.p_poly[i],poptPoly[i])
coeff = self.fitter.fit()
self.assertTrue(coeff[0] == 1.)
error = self.fitter.calcLinFitError(self.x_data)
self.assertTrue(numpy.mean(error) == 0.)
line = self.fitter.straightLine(self.x_data)
self.assertTrue(numpy.all(line - self.x_data) == 0.)
del coeff
del error
del fitPoly
del popt
del poptPoly
del fit
del line
def get_psf_fit(data, minX=0, leftY=0, verbose=False, makeplot=False):
"""
fits a pinhole (point source) continuum spectrum by a Gaussian for each row (x axis).
:Parameters:
data: MiriMeasuredModel (or similar 2-D DataModel)
The 2-D data product from which spectra will be extracted.
:Returns:
* x values per row
* fit function (vs x) per row
* all Gaussian fit parameters per row
* all errors of Gaussian fit parameters per row
"""
# Make sure the given data is a 2-D JWST data product.
if isinstance(data, MiriMeasuredModel):
if np.ndim(data.data) != 2:
strg = "lrs_extract_spec - 2 dimensional data expected "
strg += "(%d given)." % np.ndim(data.data)
raise TypeError(strg)
else:
raise TypeError("lrs_extract_spec - " + \
"Given data product is not a MiriMeasuredModel")
if np.ndim(data.err) != 2:
raise TypeError("lrs_extract_spec - " + \
"Given data product does not contain any data!")
sig = data.data
sigErr = data.err
all_x = []
all_p = []
all_fit = []
all_perr = []
for i in range(sig.shape[0]):
si = sig[i, :]
simax = np.max(si)
stdevSi = np.std(si)
if simax <= 0 or stdevSi == 0:
if verbose:
print("signal is below or equal zero in row ", i)
continue
if simax / stdevSi < 1. or simax < np.mean(sigErr):
if verbose:
print("signal too low in row ", i)
continue
maxind = np.where(si == simax)[0][0]
p_guess = (0, simax, maxind, 2)
try:
fit, p, perr = nonLinFit(gaussian,
np.arange(sig.shape[1]),
si,
y_sigma=sigErr[i, :],
p_guess=p_guess,
plotting=makeplot,
verbose=verbose)
p[2] = p[2] + leftY
all_p.append(p)
all_fit.append(fit)
all_perr.append(perr)
all_x.append(i + minX)
except:
print("fit ", i, " did not work")
return all_x, all_fit, all_p, all_perr
def lrs_extract_spec_with_fit(data,
minX=0,
leftY=0,
verbose=False,
makeplot=False):
"""
Extracts spectra from the data by summing along the spatial-x axis considering NaNs.
The row extension to sum over is determined by a Gaussian fit +- 3*FWHM
:Parameters:
data: MiriMeasuredModel (or similar 2-D DataModel)
The 2-D data product from which spectra will be extracted.
:Returns:
Spectrum1d: dataproduct with
* spec: 1d array
The extracted spectrum
* err: 1d array
The propagated error
:Raises:
TypeError
if wrong input type
"""
# Make sure the given data is a 2-D JWST data product.
if isinstance(data, MiriMeasuredModel):
if np.ndim(data.data) != 2:
strg = "lrs_extract_spec - 2 dimensional data expected "
strg += "(%d given)." % np.ndim(data.data)
raise TypeError(strg)
else:
raise TypeError("lrs_extract_spec - " + \
"Given data product is not a MiriMeasuredModel")
if np.ndim(data.err) != 2:
raise TypeError("lrs_extract_spec - " + \
"Given data product does not contain any data!")
sig = data.data
sigErr = data.err
col = []
xpixel = []
fwhm = []
for i in range(sig.shape[0]):
si = sig[i, :]
simax = np.max(si)
stdevSi = np.std(si)
if simax <= 0 or stdevSi == 0:
continue
if simax < 0.03 and simax / stdevSi < 1.:
continue
maxind = np.where(si == simax)[0][0]
p_guess = (0, simax, maxind, 2)
try:
fit, p, perr = nonLinFit(gaussian,
np.arange(sig.shape[1]),
si,
y_sigma=sigErr[i, :],
p_guess=p_guess,
plotting=False,
verbose=verbose)
col.append(p[2] + leftY)
xpixel.append(i + minX)
fwhm.append(p[3])
except:
print("fit ", i, " did not work")
col = np.asarray(col)
xpixel = np.asarray(xpixel)
fwhm = np.asarray(fwhm)
if makeplot:
mplt.plot_xy(xpixel, col, title="fit result position of peak")
stdevCol = np.std(col)
meanCol = np.mean(col)
if verbose:
print("mean column ", meanCol)
print("stddev column ", stdevCol)
colInd = np.where(
np.logical_and(col < meanCol + _sigma_clip * stdevCol,
col > meanCol - _sigma_clip * stdevCol))[0]
newXpixel = xpixel[colInd]
newCol = col[colInd]
newFwhm = fwhm[colInd]
coeff = np.polyfit(newXpixel, newCol, 1)
line = coeff[0] * newXpixel + coeff[1]
fwhmCoeff = np.polyfit(newXpixel, newFwhm, 1)
lineFwhm = fwhmCoeff[0] * newXpixel + fwhmCoeff[1]
if verbose:
print("no of useful pixels ", len(newXpixel))
if makeplot:
fig = pyplot.figure(facecolor="0.98")
fit = fig.add_subplot(211)
fit.plot(newXpixel, newCol, 'ro', np.sort(newXpixel), line)
fit.set_title("fitted determined columns")
fw = fig.add_subplot(212)
fw.plot(newXpixel, lineFwhm)
fw.set_title("FWHM")
pyplot.show()
#extract the spectra along col +- 3 * FHWM
spec = np.ndarray(sig.shape[0])
err = np.ndarray(sig.shape[0])
for i in range(sig.shape[0]):
column = coeff[0] * (i + minX) + coeff[1] - leftY
fw = fwhmCoeff[0] * (i + minX) + fwhmCoeff[1]
si = sig[i, int(column - 3. * fw):int(column + 3. * fw)]
er = sigErr[i, int(column - 3. * fw):int(column + 3. * fw)]
spec[i] = np.nansum(si)
err[i] = np.sqrt(np.nansum(np.square(er)))
if np.ndim(data.dq) > 1:
dq = data.dq
# Crunch the DQ array into 1-D by taking the maximum
# qual = np.max(dq, 1)
qual = dq.max(0)
data.dq = qual
else:
qual = None
data.data = spec
data.err = err
#spectrum = Spectrum1D(spec, error = err, quality = qual )
return data, newXpixel, line