TableModel

class sherpa.models.basic.TableModel(name='tablemodel')

Bases: ArithmeticModel

Tabulated values are linearly scaled and may be interpolated.

The load method is used to read in the tabular data. There are two different modes:

• both independent (x) and dependent (y) axes are set, in which case model evaluation will interpolate the requested grid onto the data, with the interpolation scheme controlled by the method attribute. The fold method does not need to be used.

• the x axis is set to None, in which case the grid used to evaluate the model must have the same size as the y values, and the fold method must be sent the data object to be fit before a fit is made. In this case the method attribute is ignored.

The model values - the y argument to load - are multiplied by the ampl parameter of the model.

ampl

The linear scaling factor for the table values

Notes

The model has ndim set to None as it can be used for any dimension of data. However, this only makes sense for the second mode, that is when the x argument to load is None. For the first mode, when x is not None, the model evaluation only uses the first component of the independent axes - so the low values for a Data1DInt object or the x0 values for a Data1D object.

The model ignores the integrate setting.

See also

Const1D, Scale1D

Examples

If the x array is given to the load call then the model can interpolate from the requested grid onto the load data (the default is linear interpolation):

>>> tm = TableModel()
>>> tm.load([10, 20, 25, 30], [14, 12, 17, 18])
>>> tm.ampl = 10
>>> tm([15, 20, 27])
array([130., 120., 174.])

The method attribute can be changed to select a different interpolation scheme: it requires a function that accepts (xout, xin, yin) and returns yout which are the interpolated values of xin,yin onto the grid xout:

>>> from sherpa.utils import neville
>>> tm.method = neville
>>> tm([15, 20, 27])
array([ 90.   , 120.   , 182.16])

The model can be used for data when the independent axis is either not useful - such as for an image mask - or the data does not have a meaningful independent axis, as in the the independent variable being an index for a star and the dependent axis is a property of each star. In this case the x argument to load is set to None, which means that no interpolation is used and that the fold method must be used if the data has been filtered in any way, but it’s safest to always use it:

>>> from sherpa.models.basic import TableModel
>>> from sherpa.data import Data1D
>>> from sherpa.fit import Fit
>>> d = Data1D('data', [1, 2, 3, 4, 5], [1.2, .4, 2.2, .3, 1.])
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> tm1 = TableModel('tabmodel')
>>> tm1.load(None, [.6, .2, 1.1, .2, .5])
>>> tm1.ampl.val
1.0
>>> tm1.fold(d)
>>> fit1 = Fit(d, tm1)
>>> res1 = fit1.fit()
>>> tm1.ampl.val
1.9894736842102083

In this case the fold method is necessary, to ensure that the fit only uses the valid data bins:

>>> import numpy as np
>>> y = np.ma.masked_invalid([np.nan, np.nan, 2.2, .3, 1.])
>>> d = Data1D('data', [1, 2, 3, 4, 5], y)
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> tm2 = TableModel('tabmodel')
>>> tm2.load(None, [.6, .2, 1.1, .2, .5])
>>> tm2.fold(d)
>>> tm2.ampl.val
1.0
>>> fit2 = Fit(d, tm2)
>>> res2 = fit2.fit()
>>> tm2.ampl.val
1.9866666666663104

The masking also holds if the notice or ignore method has been used on the dataset:

>>> d = Data1D('data', [1, 2, 3, 4, 5], [1.2, .4, 2.2, .3, 1])
>>> d.staterror = [.2, .2, .2, .2, .2]
>>> d.ignore(xhi=2)
>>> tm3 = TableModel('tabmodel')
>>> tm3.load(None, [.6, .2, 1.1, .2, .5])
>>> tm3.fold(d)
>>> tm3.ampl.val
1.0
>>> fit3 = Fit(d, tm3)
>>> res = fit3.fit()
>>> tm3.ampl.val
1.9866666666663104

Attributes Summary

cache	The maximum size of the cache.
lpars	Return any linked parameters.
method	The interpolation method, used when x is not None in the load call.
ndim	The dimensionality of the model, if defined, or None.
pars	Return the parameters of the model.
thawedparhardmaxes	The hard maximum values for the thawed parameters.
thawedparhardmins	The hard minimum values for the thawed parameters.
thawedparmaxes	The maximum limits of the thawed parameters.
thawedparmins	The minimum limits of the thawed parameters.
thawedpars	The thawed parameters of the model.

Methods Summary

apply(outer, *otherargs, **otherkwargs)
cache_clear()	Clear the cache.
cache_status()	Display the cache status.
calc(p, x0[, x1])	Evaluate the model.
fold(data)	Ensure the model matches the data filter.
freeze()	Freeze any thawed parameters of the model.
get_center()
get_thawed_pars()	Return the thawed parameter objects.
get_x()	Return the independent axis or None.
get_y()	Return the dependent axis or None.
guess(dep, *args, **kwargs)	Set an initial guess for the parameter values.
load(x, y)	Set the model values.
reset()	Reset the parameter values.
set_center(*args, **kwargs)
startup([cache])	Called before a model may be evaluated multiple times.
teardown()	Called after a model may be evaluated multiple times.
thaw()	Thaw any frozen parameters of the model.

Attributes Documentation

cache = 5

The maximum size of the cache.

lpars

Return any linked parameters.

This only returns linked parameters that are not related to the model, and each parameter is not repeated.

Added in version 4.16.1.

See also

pars

Examples

By default there are no linked parameters:

>>> from sherpa.models.basic import Gauss2D
>>> mdl = Gauss2D("mdl")
>>> len(mdl.pars)
6
>>> mdl.lpars
()

Force the model to have identical xpos and ypos parameters. Since the linked parameter value (mdl.xpos) is part of the model it is not included in lpars:

>>> mdl.ypos = mdl.xpos
>>> len(mdl.pars)
6
>>> mdl.lpars
()

Add a link to allow the sigma term to be fit rather than FWHM. Since the linked parameter - here from the Const1D model - is not a part of the model it is included in lpars:

>>> import numpy as np
>>> from sherpa.models.basic import Const1D
>>> sigma = Const1D("sigma")
>>> mdl.fwhm = 2 * np.sqrt(2 * np.log(2)) * sigma.c0
>>> len(mdl.pars)
6
>>> mdl.lpars
(<Parameter 'c0' of model 'sigma'>,)

method

The interpolation method, used when x is not None in the load call.

The method argument is a function that accepts arguments (xout, xin, yin) and returns the yout values from interpolating xout onto (xin, yin). The default is linear interpolation (sherpa.utils.linear_interp).

ndim: int | None = None

The dimensionality of the model, if defined, or None.

pars

Return the parameters of the model.

This does not include any linked parameters.

Changed in version 4.16.1: The pars field can no-longer be set directly. Individual elements can still be changed.

See also

lpars

thawedparhardmaxes

The hard maximum values for the thawed parameters.

The minimum and maximum range of the parameters can be changed with thawedparmins and thawedparmaxes but only within the range given by thawedparhardmins to thawparhardmaxes.

See also

thawedparhardmins, thawedparmaxes

thawedparhardmins

The hard minimum values for the thawed parameters.

The minimum and maximum range of the parameters can be changed with thawedparmins and thawedparmaxes but only within the range given by thawedparhardmins to thawparhardmaxes.

See also

thawedparhardmaxes, thawedparmins

thawedparmaxes

The maximum limits of the thawed parameters.

Get or set the maximum limits of the thawed parameters of the model as a list of numbers. If there are no thawed parameters then [] is used. The ordering matches that of the pars attribute.

See also

thawedpars, thawedarhardmaxes, thawedparmins

thawedparmins

The minimum limits of the thawed parameters.

Get or set the minimum limits of the thawed parameters of the model as a list of numbers. If there are no thawed parameters then [] is used. The ordering matches that of the pars attribute.

See also

thawedpars, thawedarhardmins, thawedparmaxes

thawedpars

The thawed parameters of the model.

Get or set the thawed parameters of the model as a list of numbers. If there are no thawed parameters then [] is used. The ordering matches that of the pars attribute.

See also

thawedparmaxes, thawedparmins

Methods Documentation

apply(outer, *otherargs, **otherkwargs)

cache_clear() → None

Clear the cache.

cache_status() → None

Display the cache status.

Information on the cache - the number of “hits”, “misses”, and “requests” - is displayed at the INFO logging level.

Example

>>> pl.cache_status()
 powlaw1d.pl                size:    5  hits:   633  misses:   240  check=  873

calc(p, x0, x1=None, *args, **kwargs)

Evaluate the model.

The load method must have been called first. If both x and y were given then the model is interpolated onto the x0 grid, otherwise x0 is only checked to see if it has the right size (fold should have been called if the data has been filtered).

fold(data)

Ensure the model matches the data filter.

This should be called after load, to ensure that any existing filter is applied correctly during a fit. It is only necessary for models where x is None in the load call.

Parameters:

data (sherpa.data.Data instance) – An object with a mask attribute.

freeze() → None

Freeze any thawed parameters of the model.

get_center()

get_thawed_pars() → list[Parameter]

Return the thawed parameter objects.

This includes linked parameters, which complicates the min/max settings, since the range on the components of a linked parameter does not match that of the original parameter, which is an issue when the limits are exceeded.

Added in version 4.16.1.

get_x()

Return the independent axis or None.

See also

get_y, load

get_y()

Return the dependent axis or None.

See also

get_x, load

guess(dep, *args, **kwargs)

Set an initial guess for the parameter values.

Attempt to set the parameter values, and ranges, for the model to match the data values. This is intended as a rough guess, so it is expected that the model is only evaluated a small number of times, if at all.

load(x, y)

Set the model values.

Parameters:

• x (None or sequence) – The model values. It is expected that either both are given, and have the same number of elements, or that only y is set, although the model can be cleared by setting both to None. If x is given then the data is saved after being sorted into increasing order of x.

• y (None or sequence) – The model values. It is expected that either both are given, and have the same number of elements, or that only y is set, although the model can be cleared by setting both to None. If x is given then the data is saved after being sorted into increasing order of x.

See also

get_x, get_y

reset() → None

Reset the parameter values.

Restores each parameter to the last value it was set to. This allows the parameters to be easily reset after a fit.

set_center(*args, **kwargs)

startup(cache: bool = False) → None

Called before a model may be evaluated multiple times.

Parameters:

cache (bool, optional) – Should a cache be used when evaluating the models.

See also

teardown

teardown() → None

Called after a model may be evaluated multiple times.

See also

startup

thaw() → None

Thaw any frozen parameters of the model.

Those parameters that are marked as “always frozen” are skipped.