Last modified: December 2023

Jump to: Description · Examples · PARAMETERS · Notes · Bugs · See Also

AHELP for CIAO 4.16 Sherpa


Context: confidence


Plot the statistic value as two parameters are varied.


reg_proj(par0, par1, id=None, otherids=None, replot=False, fast=True,
min=None, max=None, nloop=(10, 10), delv=None, fac=4, log=(False,
False), sigma=(1, 2, 3), levels=None, numcores=None, overplot=False)

id - int or str, optional
otherids - sequence of int or str, optional
replot - bool, optional
fast - bool, optional
min - pair of numbers, optional
max - pair of number, optional
nloop - pair of int, optional
delv - pair of number, optional
fac - number, optional
log - pair of bool, optional
sigma - sequence of number, optional
levels - sequence of number, optional
numcores - optional
overplot - bool, optional


Create a confidence plot of the fit statistic as a function of parameter value. Dashed lines are added to indicate the current statistic value and the parameter value at this point. The parameter value is varied over a grid of points and the free parameters re-fit. It is expected that this is run after a successful fit, so that the parameter values are at the best-fit location.


Example 1

Vary the xpos and ypos parameters of the gsrc model component for all data sets with a source expression.

>>> reg_proj(gsrc.xpos, gsrc.ypos)

Example 2

Use only the data in data set 1:

>>> reg_proj(gsrc.xpos, gsrc.ypos, id=1)

Example 3

Only display the one- and three-sigma contours:

>>> reg_proj(gsrc.xpos, gsrc.ypos, sigma=(1, 3))

Example 4

Display contours at values of 5, 10, and 20 more than the statistic value of the source model for data set 1:

>>> s0 = calc_stat(id=1)
>>> lvls = s0 + np.asarray([5, 10, 20])
>>> reg_proj(gsrc.xpos, gsrc.ypos, levels=lvls, id=1)

Example 5

Increase the limits of the plot and the number of steps along each axis:

>>> reg_proj(gsrc.xpos, gsrc.ypos, id=1, fac=6, nloop=(41, 41))

Example 6

Compare the ampl parameters of the g and b model components, for data sets 'core' and 'jet', over the given ranges:

>>> reg_proj(g.ampl, b.ampl, min=(0, 1e-4), max=(0.2, 5e-4),
...          nloop=(51, 51), id='core', otherids=['jet'])


The parameters for this function are:

Parameter Definition
par0 The parameters to plot on the X and Y axes, respectively.
par1 The parameters to plot on the X and Y axes, respectively.
id The data set that provides the data. If not given then all data sets with an associated model are used simultaneously.
otherids Other data sets to use in the calculation.
replot Set to True to use the values calculated by the last call to `int_proj` . The default is False .
fast If True then the fit optimization used may be changed from the current setting (only for the error analysis) to use a faster optimization method. The default is False .
min The minimum parameter value for the calculation. The default value of none means that the limit is calculated from the covariance, using the `fac` value.
max The maximum parameter value for the calculation. The default value of none means that the limit is calculated from the covariance, using the `fac` value.
nloop The number of steps to use. This is used when `delv` is set to none .
delv The step size for the parameter. Setting this over-rides the `nloop` parameter. The default is none .
fac When `min` or `max` is not given, multiply the covariance of the parameter by this value to calculate the limit (which is then added or subtracted to the parameter value, as required).
log Should the step size be logarithmically spaced? The default ( False ) is to use a linear grid.
sigma The levels at which to draw the contours. The units are the change in significance relative to the starting value, in units of sigma.
levels The numeric values at which to draw the contours. This over-rides the `sigma` parameter, if set (the default is none ).
numcores The number of CPU cores to use. The default is to use all the cores on the machine.
overplot If True then add the data to an existing plot, otherwise create a new plot. The default is False .


The difference to `reg_unc` is that at each step, a fit is made to the remaining thawed parameters in the source model. This makes the result a more-accurate rendering of the projected shape of the hypersurface formed by the statistic, but the run-time is longer than, the results of `reg_unc` , which does not vary any other parameter. If there are no free parameters in the model, other than the parameters being plotted, then the results will be the same.


See the bugs pages on the Sherpa website for an up-to-date listing of known bugs.

See Also

conf, confidence, covar, covariance, get_conf, get_conf_results, get_covar, get_covar_opt, get_covar_results, get_covariance_results, get_int_proj, get_int_unc, get_proj, get_proj_opt, get_proj_results, get_projection_results, get_reg_proj, get_reg_unc, int_proj, int_unc, proj, projection, reg_unc, set_conf_opt, set_covar_opt, set_proj_opt