Tutorial: Exploring Cube Fits

Having fit our cube, we can pull out a bunch of parameters and have a look at the data. Following the previous tutorial, we have an hf_fitter object, that contains the maps, which is a dictionary that we can pull things out of:

maps = hf_fitter.parameter_maps

# Get the reduced chi-square if it's a numpy array
chisq_red = maps["chisq_red"]

# Or if you've input a fits file to start with, then this will be a fits image
chisq_red = maps["chisq_red"].data

You can then plot this up however you might like:

The fitted parameters (and errors) are also in here as [param]_[comp_no], [param]_[comp_no]_err_up, and [param]_[comp_no]_err_down. Note these don’t take into account the covariances, if you want to properly sample for the purposes of a plot then you’ll need the full set of walkers, which can be saved in individual fit dictionary files.

You can also plot things in much the same way as in the single spectra plotting way, except you can also provide a grid to cut down the number of plots created. For example:

plot_dir = "plots"

grid = np.zeros_like(chisq_red)

grid[10, 10] = 1
grid[50, 50] = 1

if not os.path.exists(plot_dir):
    os.makedirs(plot_dir)

mcfine_fitting_filename = os.path.join(
        fit_dir, f"mcfine_fit_backward",
    )

hf_plotter.plot_step(plot_name=os.path.join(fit_dir, plot_dir, f"{target}_step"),
                     mcfine_output_filename=mcfine_output_filename,
                     #fit_dict_filename=os.path.join(coherence_backward_dir, fit_dict_filename),
                     n_comp_filename=os.path.join(coherence_backward_dir, n_comp_filename),
                     grid=grid)
hf_plotter.plot_corner(plot_name=os.path.join(plot_dir, f"{target}_corner"),
                       mcfine_output_filename=mcfine_output_filename,
                       #fit_dict_filename=os.path.join(coherence_backward_dir, fit_dict_filename),
                       grid=grid)
hf_plotter.plot_fit(plot_name=os.path.join(plot_dir, f"{target}"),
                    mcfine_output_filename=mcfine_output_filename,
                    #fit_dict_filename=os.path.join(coherence_backward_dir, fit_dict_filename),
                    grid=grid)

Note here that you can either pass the full fit dictionary (mcfine_output_filename), or each individual fit dictionary (fit_dict_filename).