def line_fit(p, x, y, y_error, x_error=False, iterations=10000):
"""
Linear Fit to data, taking either errors in y or both in x and y into
account.
Parameters
----------
p : list
Containg slope (m) and y-axis intersection (b) p=[m, b]. Same as in
:func:`line` and :func:`antiline`.
x : float or list
x measurements. Data.
y : float or list
y measurements. Data.
y_error : float or list
The y measurment errors.
x_error : float or list
The x measurment errors. If unset only errors in y are taken into
account.
"""
p1, cov, info, mesg, success = least(linear_error_function, p,
args=(x, y, y_error, x_error),
maxfev=int(iterations), full_output=1)
dof = len(x) - len(p) - 1 # degrees of Freedom
chisq = sum(info['fvec'] * info['fvec']) / dof
Error = std(info['fvec'])
return p1, chisq, Error
def _grid_fit(data, beta, nu_or_lambda='nu', fit_beta=False, rawChiSq=False,
kappa='Kruegel'):
r"""
Different approach to find the best fitting SED using certain ranges of
temperatures and masses to avoid unreasonably high values.
Not sure about the functionality and the code is badly written with the
temperature and mass ranges hard coded. Thats why its not public.
Once approved it may be made public again.
Parameters
----------
data : array
Same format as in grey_body_fit.
beta :
The beta value. If fit_beta=True this is varied in the fits.
kappa :
As in :func:`greybody`.
Other Parameters
----------------
fit_beta: True or False
Steers if beta is fittet or not.
rawChisq: True or False
Returns chi square without normalisation, or not.
"""
def _err(p, x, y, y_error, nu_or_lambda):
"""
The function to be minimized by scipy.optimize.leastsq. It returns the
difference between the measured fluxes, y, and the calculated fluxes
for the parameters p of the SED at the given frequency x.
Parameters
----------
p : list
same start start_parameter as in grey_body_fit
x and y :
The two rows, data[0] and data[1] of the data variable from
grey_body_fit.
y_error : The absolute error on the flux measurement.
Other Parameters
----------------
nu_or_lambda : string
Specify whether x is a frequency :math:`\nu` ``'nu'`` or
a wavelenght :math:`\lambda` ``'lambda'``; default is ``'nu'``.::
**Don't** use ``'lambda'`` as this part of the
:py:func:`grey_body` is not up-to-date.
Notes
-----
This function caluclates the residuals that are minimized by
:func:`leastsq`, thus it calculates:
.. math::
(model-measurement)/\sigma
Note the :math:`\chi^2` is defined as:
.. math::
\chi^2 = \frac{1}{N} \sum{(model-measurement)/\sigma}
:warning:`Formula has to be checked.`
This functions is different from the _err functions in grey_body_fit
because the start parameters are handles differently.
"""
if not fit_beta:
pMulti = [T, p, beta]
print pMulti
if fit_beta:
pMulti = [Temps, p]
print pMulti
return (((multi_component_grey_body(pMulti, x, nu_or_lambda, kappa)[0])
- y) / y_error)
beta = [beta]
T1 = arange(5, 70, 1)
T2 = arange(20, 100, 1)
chisqList = {}
chisqList1 = []
for i in T1:
for j in T2:
T = [i, j]
N = [1e2, 1]
if not fit_beta:
p = N
if fit_beta:
p = N + beta
p2, cov, info, mesg, success = least(_err, p, args=(data[0],
data[1], data[2],
nu_or_lambda),
maxfev=int(1e9),
full_output=1)
dof = len(data[0]) - len(p) - 1 # Degrees of Freedom
chisq = sum(info['fvec'] * info['fvec']) / dof # see above
rawchisq = sum(info['fvec'] * info['fvec'])
chisqList[chisq] = [[i, j], p2]
chisqList1 += [chisq]
chisq = sorted(chisqList)[0]
T, p2 = chisqList[sorted(chisqList)[0]]
if not fit_beta:
numberOfComponents = len(p)
p2 = [T, p2, beta]
if fit_beta:
numberOfComponents = (len(p) - 1)
p2 = [Temps, list(p2[0:numberOfComponents]), [p2[len(p) - 1]]]
if not rawChiSq:
return p2, chisq
if rawChiSq:
return p2, rawchisq
return sorted(chisq)[0]
def grey_body_fit(data, start_parameter, nu_or_lambda='nu', fit_beta=False,
fix_temperature=False, rawChiSq=None, kappa='Kruegel',
residuals=False, iterations=1e9):
r"""
This function fits a multi component grey body model to an observed SED for
the optical thin case.
Parameters
----------
data : array
The obseved data. Array of shape(3, x) first row has to be the X values
(Frequency in [GHz]) of the measurements, second row the Y values
(Flux [Jy]), and the third row the Z values the errors on the
fluxes i.e.:
data = array([[X1, X2, X3, ...], [Y1, Y2, Y3,...], [Z1, Z2,
Z3, ...]])
start_parameter : array
Array of a first guess of the parameters of the grey_body components.
The number of components is arbitrary.
start_parameter = [[T1, T2, T3,...], [N1, N2, N3, ...], beta]
fit_beta : True or False
If True Beta is allowed to vary. Default is False.
fix_temperature : True or False
If True the Temperature is fixed allowed to vary.
rawChiSq :
if None the function gives the reduced chisq Value. If True the
function gives chisq without dividing it by the dof
Returns
-------
p2 : list
The final grey_body parameters that reduce the least squares for the
given dataset.
chisq/rawChiSq :
chisq is reduced chisq with degrees of freedom:
dof= #dataPoints-#freeFitParameters-1
Other Parameters
----------------
nu_or_lambda : string
Specify whether x is a frequency :math:`\nu` ``'nu'`` or
a wavelenght :math:`\lambda` ``'lambda'``; default is ``'nu'``.::
**Don't** use ``'lambda'`` as this part of the
:py:func:`grey_body` is not up-to-date.
See Also
--------
scipy.optimize.leastsq: This function is used to perform the least squares
fit.
multi_component_grey_body, grey_body, black_body: Defining the function to
be fittet to the data.
Notes
-----
A one component fit has four free parameters if beta is allowed to vary or
three if beta is fixed (one more than parameters to fit). Each additional
component adds two more free paramters to fit.
Assure that:
number of data points > number of free parameters.
"""
# Distinguish between the different options and build the parameter list
# needed by optimize.leastsq()
if not fit_beta:
if not fix_temperature:
p = start_parameter[0] + start_parameter[1]
beta = start_parameter[2]
if fix_temperature:
p = start_parameter[1]
Temps = start_parameter[0]
beta = start_parameter[2]
if fit_beta:
if not fix_temperature:
p = start_parameter[0] + start_parameter[1] + start_parameter[2]
if fix_temperature:
p = start_parameter[1] + start_parameter[2]
Temps = start_parameter[0]
def _err(p, x, y, y_error, nu_or_lambda):
"""
The function to be minimized by scipy.optimize.leastsq. It returns the
difference between the measured fluxes, y, and the calculated fluxes
for the parameters p of the SED at the given frequency x.
Parameters
----------
p : list
same start start_parameter as in grey_body_fit
x and y :
The two rows, data[0] and data[1] of the data variable from
grey_body_fit.
y_error : The absolute error on the flux measurement.
Other Parameters
----------------
nu_or_lambda : string
Specify whether x is a frequency :math:`\nu` ``'nu'`` or
a wavelenght :math:`\lambda` ``'lambda'``; default is ``'nu'``.::
**Don't** use ``'lambda'`` as this part of the
:py:func:`grey_body` is not up-to-date.
Notes
-----
This function caluclates the residuals that are minimized by
:func:`leastsq`, thus it calculates:
.. math::
(model-measurement)/\sigma
Note the :math:`\chi^2` is defined as:
.. math::
\chi^2 = \frac{1}{N} \sum{(model-measurement)/\sigma}
:warning:`Formula has to be checked.`
"""
if not fit_beta:
if not fix_temperature:
numberOfComponents = (len(p)) / 2
pMulti = [list(p[0:numberOfComponents]),
list(p[numberOfComponents:numberOfComponents * 2])]
pMulti += [beta]
if fix_temperature:
pMulti = [Temps, p, beta]
if fit_beta:
if not fix_temperature:
numberOfComponents = (len(p) - 1) / 2
pMulti = [list(p[0:numberOfComponents]),
list(p[numberOfComponents:numberOfComponents * 2]),
p[len(p) - 1]]
if fix_temperature:
numberOfComponents = len(p) - 1
pMulti = [Temps, list(p[0:numberOfComponents]), p[len(p) - 1]]
return (((multi_component_grey_body(pMulti, x, nu_or_lambda,
kappa=kappa)[0]) - y) / y_error)
# The actual fit
# maxfev : Number of integrations
# full_output: Additional informations
# args: X and Y data
p2, cov, info, mesg, success = least(_err, p, args=(data[0], data[1],
data[2], nu_or_lambda),
maxfev=int(iterations), full_output=1)
# return of the optimized parameters
dof = len(data[0]) - len(p) - 1 # degrees of Freedom
chisq = sum(info['fvec'] * info['fvec']) / dof # see above
rawchisq = sum(info['fvec'] * info['fvec'])
if not fit_beta:
if not fix_temperature:
numberOfComponents = (len(p)) / 2
p2 = [list(p2[0:numberOfComponents]),
list(p2[numberOfComponents:numberOfComponents * 2]), beta]
if fix_temperature:
numberOfComponents = len(p)
p2 = [Temps, p2, beta]
if fit_beta:
if not fix_temperature:
numberOfComponents = (len(p) - 1) / 2
p2 = [list(p2[0:numberOfComponents]),
list(p2[numberOfComponents:numberOfComponents * 2]),
[p2[len(p) - 1]]]
if fix_temperature:
numberOfComponents = (len(p) - 1)
p2 = [Temps, list(p2[0:numberOfComponents]), [p2[len(p) - 1]]]
if residuals:
return p2, chisq, info['fvec']
if rawChiSq == None:
return p2, chisq
if rawChiSq:
return p2, rawchisq