Analysing and Plotting Multiple Models

A key feature of the sage-analysis package is its ability to easily and succinctly analyse and plotting a number of different SAGE Models. This makes it easy to compare different models very quickly.

We have covered the basics of plotting a single model here. In this example, we touch upon those things that must change to visualize multiple models.

Setting Things Up

Again, specify what you want to plot in an identical manner to previously.

# Going to just plot the stellar mass function.
plot_toggles = {"SMF": 1}

plot_output_format = "png"
plot_output_path = "./plots"

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

Defining the Model Dictionaries

Each Model requires its own dictionary to define its input files and other parameters. For this example, let’s suppose we wish to compare the results of running SAGE on the default Mini-Millennium simulation and Mini-Millennium without supernova feedback.

millennium = { "snapshot": 63,   # Snapshot we're plotting properties at.
               "IMF": "Chabrier",  # Chabrier or Salpeter.
               "label": "Mini-Millennium",  # Legend label.
               "sage_file": "../input/millennium.par",
               "sage_output_format": "sage_hdf5",
               "first_file": 0,  # File range (or core range for HDF5) to plot.
               "last_file": 0,  # Closed interval, [first_file, last_file].
             }

millennium_noSN = { "snapshot": 63,   # Snapshot we're plotting properties at.
                    "IMF": "Chabrier",  # Chabrier or Salpeter.
                    "label": "Mini-Millennium-NoSN",  # Legend label.
                    "sage_file": "../input/millennium_noSN.par",
                    "sage_output_format": "sage_binary",
                    "first_file": 0,  # File range (or core range for HDF5) to plot.
                    "last_file": 0,  # Closed interval, [first_file, last_file].
                    "num_output_files": 1,
             }

sims_to_plot = [millennium, millennium_noSN]
models = []

The important line here is the very last line. We place all the models we wish to analyse into a single list, sims_to_plot. Then, in the following code block, we will iterate through each of these simulations and place the Model classes in the model list.

NOTE: For example purposes, we use the binary SAGE output option for the no supernovae simulation. Hence, we must specify the number of output files that SAGE wrote to, num_output_files.

Iterating Through Models

Analysing multiple models is down in an identical manner as shown previously. The only difference is that we wrap the code in a for loop and iterate through models_to_plot. We defer to the previous page for a full explanation of the following code block.

# Set up the dictionaries for the calculation and plotting functions.
from sage_analysis.utils import generate_func_dict

# Search for functions named "calc_<plot_toggle>" in the "example_calcs"
# module and "plot_<plot_toggle>" in the "example_plots" module.
calculation_functions = generate_func_dict(
                            plot_toggles,
                            module_name="sage_analysis.example_calcs",
                            function_prefix="calc"
                            )

plot_functions = generate_func_dict(
                    plot_toggles,
                    module_name="sage_analysis.example_plots",
                    function_prefix="plot_"
                    )

# Iterate through the simulations and set up a Model for each.
from sage_analysis.model import Model
from sage_analysis.sage_hdf5 import SageHdf5Data
from sage_analysis.sage_binary import SageBinaryData

for model_dict in sims_to_plot:
    model = Model()
    model.plot_output_format = plot_output_format

    # This switch case should be extended if you're defining your own
    # custom data class.
    if model_dict["sage_output_format"] == "sage_hdf5":
        model.data_class = SageHdf5Data(model, millennium["sage_file"])
    elif model_dict["sage_output_format"] == "sage_binary":
        model.data_class = SageBinaryData(model, model_dict["num_output_files"],
                                          model_dict["sage_file"],
                                          model_dict["snapshot"])
    else:
        raise ValueError

    # The data class has read the SAGE ini file.  Update the model with the parameters
    # read and those specified by the user.
    model.update_attributes(model_dict)

    # Initialize the properties for this model. Only plotting the SMF.
    stellar_properties = ["SMF", "red_SMF", "blue_SMF"]
    min_mass = 8.0  # log10(Msun).
    max_mass = 12.0  # log10(Msun).
    bin_width = 0.1  # log10(Msun).
    bin_name = "stellar_mass_bins"
    model.init_binned_properties(min_mass, max_mass, bin_width, bin_name,
                                 stellar_properties)

    # Calculate all the properties.
    model.calc_properties_all_files(calculations_functions)

    # Append this model to the list.
    models.append(model)

Plotting Multiple Models

All the hard work has already been done! The plotting functions defined in plot_functions accept a list of models as their first argument. Hence, lets just pass in models!

for func_name in plot_functions.keys():
    func = plot_functions[func_name][0]
    func(models, plot_output_path, plot_output_format)

This produces the stellar mass function for multiple models. From this, we can better understand the role that supernova feedback plays in regulating galaxy growth.

Using Keyword Arguments

Both the calculation and plotting functions support the use of optional keyword arguments. This allows finer control over plotting (e.g.) galaxy sub-populations. The procedure for using this option for multiple models identical to the single Model scenario which we show in Using Keyword Arguments.

Analysing Multiple Models Over Redshift

To analyse multiple models over a number of redshifts, one simply needs to wrap the code outlined here in the for model in sim_to_plot loop.

We defer to the history module for a full example of analysing multiple models over redshifts.