Non-Square Matrix Plots

This example demonstrates how to create a non-square matrix of correlation plots. This is useful for comparing a single application (or perhaps a few) with a suite of experiments (like those from the ICSBEP handbook [Bess2019]).

Generating the Plot

The steps for creating the matrix plot are just as described in Matrix of Contribution Correlation Plots except that the application files and experiment files lists are not the same length. Recall, first the uncertainty contributions must be generated from the given SDF profiles using tsunami_ip_utils.integral_indices.get_uncertainty_contributions(). In this example, we will choose a single application from the MCT directory, and compare it to all other SDFS in the MCT directory.

from tsunami_ip_utils.viz.viz import matrix_plot
from tsunami_ip_utils.viz.plot_utils import generate_plot_objects_array_from_contributions
from tsunami_ip_utils.integral_indices import get_uncertainty_contributions
from paths import EXAMPLES
import os

# Get the filenames of all of the SDF profiles in the MCT directory
all_mct_sdfs = os.listdir(EXAMPLES / 'data' / 'example_sdfs' / 'MCT')

# Generate the lists of application and experiment SDF file paths
application_files = [ EXAMPLES / 'data' / 'example_sdfs' / 'MCT' / f'MIX-COMP-THERM-001-001.sdf' ]
experiment_files = [ EXAMPLES / 'data' / 'example_sdfs' / 'MCT' / filename for filename in all_mct_sdfs
                    if filename != 'MIX-COMP-THERM-001-001.sdf' ]

contributions_nuclide, contributions_nuclide_reaction = get_uncertainty_contributions(application_files, experiment_files, variance=True)

Then the matrix of plot objects (that is plotted using tsunami_ip_utils.viz.viz.matrix_plot()) can be generated using tsunami_ip_utils.viz.plot_utils.generate_plot_objects_array_from_contributions() (note we also label the matrix cells by the application and experiment sdf filenames by explicitly passing the labels dictionary as an argument).

labels = {
    'applications': [ application_file.name for application_file in application_files ],
    'experiments': [ experiment_file.name for experiment_file in experiment_files ],
}

plot_objects_array = generate_plot_objects_array_from_contributions(
    contributions_nuclide,
    integral_index_name='(Δk/k)^2'
)

Finally, the matrix plot can be generated and displayed.

fig = matrix_plot(plot_objects_array, plot_type='interactive', labels=labels)
fig.show()

- Zoom Out + Zoom In

Sorting the Matrix

If you’re comparins a single application to multiple experiments, it may be useful to sort the experiments by the computed correlation coefficient (pearson, spearman, etc.). Since a plot objects array is first generated, then plotted using the tsunami_ip_utils.viz.viz.matrix_plot() function, we may arbitrarily reorder the plot objects array before plotting. However, we must be careful to reorder the labels dictionary in the same way as the plot objects array. The following code demonstrates how to sort the experiments by the pearson correlation coefficient.

import numpy as np

# Sort the plot objects array by the pearson correlation coefficient using numpy.argsort and get the indices representing the sorted
# ordering of plot objects.

sorted_indices = np.argsort([ plot_object.statistics['pearson'] for plot_object in plot_objects_array[:, 0] ])

# Note [:,0] indexes the fist column of the matrix plot, i.e. the column corresponding to the first application.

Note that by default numpy.argsort() returns the indices that would sort the array in ascending order. If you want to sort in descending order, you can reverse the order of the indices as follows:

sorted_indices = sorted_indices[::-1]

# Sort the plot objects by the indices
sorted_plot_objects_array = plot_objects_array[sorted_indices]

# Sort the experiment labels by the indices
labels['experiments'] = [ labels['experiments'][index] for index in sorted_indices ]

# now plot the sorted matrix
fig = matrix_plot(sorted_plot_objects_array, plot_type='interactive', labels=labels)
fig.show()

- Zoom Out + Zoom In

A static image of the plot can be saved using the tsunami_ip_utils.viz.matrix_plot.InteractiveMatrixPlot.to_image() method

fig = matrix_plot(plot_objects_array[:4, :], plot_type='interactive') # Note we're only saving a portion of the matrix to get a nice image
fig.to_image( EXAMPLES / '_static' / 'nonsquare_matrix.png' )

# sphinx_gallery_thumbnail_path = '../../examples/_static/nonsquare_matrix.png'

- Zoom Out + Zoom In