New Visualizations

import hddm
from matplotlib import pyplot as plt
import numpy as np
from copy import deepcopy

Generate some Data

# Metadata
nmcmc = 2000
model = 'angle'
n_samples = 1000
includes = hddm.simulators.model_config[model]['hddm_include']

data, full_parameter_dict = hddm.simulators.hddm_dataset_generators.simulator_h_c(n_subjects = 5,
                                                                                  n_trials_per_subject = n_samples,
                                                                                  model = model,
                                                                                  p_outlier = 0.00,
                                                                                  conditions = None,
                                                                                  depends_on = None,
                                                                                  regression_models = None,
                                                                                  regression_covariates = None,
                                                                                  group_only_regressors = False,
                                                                                  group_only = None,
                                                                                  fixed_at_default = None)

data
rt response subj_idx v a z t theta
0 1.590851 0.0 0 -1.704336 0.917225 0.612362 0.898857 0.251476
1 1.784849 0.0 0 -1.704336 0.917225 0.612362 0.898857 0.251476
2 1.927849 0.0 0 -1.704336 0.917225 0.612362 0.898857 0.251476
3 1.390854 0.0 0 -1.704336 0.917225 0.612362 0.898857 0.251476
4 1.839848 0.0 0 -1.704336 0.917225 0.612362 0.898857 0.251476
... ... ... ... ... ... ... ... ...
4995 1.441180 0.0 4 -1.314454 1.491856 0.488526 0.915183 0.119143
4996 1.239182 0.0 4 -1.314454 1.491856 0.488526 0.915183 0.119143
4997 1.708176 0.0 4 -1.314454 1.491856 0.488526 0.915183 0.119143
4998 1.447180 0.0 4 -1.314454 1.491856 0.488526 0.915183 0.119143
4999 1.729176 0.0 4 -1.314454 1.491856 0.488526 0.915183 0.119143

5000 rows × 8 columns

# Define the HDDM model
hddmnn_model = hddm.HDDMnn(data,
                           informative = False,
                           include = includes,
                           p_outlier = 0.0,
                           model = model)

# Sample
hddmnn_model.sample(nmcmc,
                    burn = 100)

[-----------------100%-----------------] 2000 of 2000 complete in 450.8 sec

<pymc.MCMC.MCMC at 0x14342cdd0>

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (12, 8),
                                        value_range = np.arange(0, 3, 0.1),
                                        plot_func = hddm.plotting._plot_func_model,
                                        parameter_recovery_mode = True,
                                        **{'add_legend': False,
                                           'alpha': 0.01,
                                           'ylim': 6.0,
                                           'bin_size': 0.025,
                                           'add_posterior_mean_model': True,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_model': True,
                                           'add_posterior_uncertainty_rts': False,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'legend_loc': 'upper left',
                                           'linewidth_histogram': 1.0,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.35, 'wspace': 0.3}})
_images/lan_visualizations_8_0.png

caterpillar plot

The caterpillar_plot() function below displays parameterwise,

  • as a blue tick-mark the ground truth.

  • as a thin black line the \(1 - 99\) percentile range of the posterior distribution

  • as a thick black line the \(5-95\) percentile range of the posterior distribution

Again use the help() function to learn more.

# Caterpillar Plot: (Parameters recovered ok?)
hddm.plotting.plot_caterpillar(hddm_model = hddmnn_model,
                               ground_truth_parameter_dict = full_parameter_dict,
                               drop_sd = False,
                               figsize = (8, 6))
_images/lan_visualizations_11_0.png

posterior predictive (standard)

We have two versions of the standard posterior predictive plots. Both are called via the plot_posterior_predictive function, however you pick the plot_func argument accordingly.

If you pick hddm.plotting.plot_func_posterior_pdf_node_nn, the resulting plot uses likelihood pdf evaluations to show you the posterior predictive including uncertainty.

If you pick hddm.plotting.plot_func_posterior_pdf_node_from_sim, the posterior predictives instead derive from actual simulation runs from the model.

Both function have a number of options to customize styling.

Below we illustrate:

_plot_func_posterior_pdf_node_nn

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (8, 6),
                                        value_range = np.arange(-4, 4, 0.01),
                                        plot_func = hddm.plotting._plot_func_posterior_pdf_node_nn,
                                        parameter_recovery_mode = True,
                                        **{'alpha': 0.01,
                                           'ylim': 3,
                                           'bin_size': 0.05,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_rts': True,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.3, 'wspace': 0.3}})
plt.show()
_images/lan_visualizations_15_0.png

_plot_func_posterior_node_from_sim

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (8, 6),
                                        value_range = np.arange(-6, 6, 0.02),
                                        plot_func = hddm.plotting._plot_func_posterior_node_from_sim,
                                        parameter_recovery_mode = True,
                                        **{'alpha': 0.01,
                                           'ylim': 3,
                                           'bin_size': 0.1,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_rts': False,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.3, 'wspace': 0.3}})
_images/lan_visualizations_17_0.png

posterior predictive (model cartoon)

The model plot is useful to illustrate the behavior of a models pictorially, including the uncertainty over model parameters embedded in the posterior distribution.

This plot works only for 2-choice models at this point.

Check out more of it’s capabilities with the help() function.

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (8, 6),
                                        value_range = np.arange(0, 4, 0.1),
                                        plot_func = hddm.plotting._plot_func_model,
                                        parameter_recovery_mode = True,
                                        **{'alpha': 0.01,
                                           'ylim': 3,
                                           'add_posterior_mean_model': True,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_model': True,
                                           'add_posterior_uncertainty_rts': True,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.3, 'wspace': 0.3}})
plt.show()
_images/lan_visualizations_20_0.png

posterior pair plot

hddm.plotting.plot_posterior_pair(hddmnn_model, save = False,
                                  parameter_recovery_mode = True,
                                  samples = 200,
                                  figsize = (6, 6))
_images/lan_visualizations_22_0.png _images/lan_visualizations_22_1.png _images/lan_visualizations_22_2.png _images/lan_visualizations_22_3.png _images/lan_visualizations_22_4.png

NEW (v0.9.3): Plots for \(n>2\) choice models

# Metadata
nmcmc = 1000
model = 'race_no_bias_angle_4'
n_samples = 1000
includes = deepcopy(hddm.simulators.model_config[model]['hddm_include'])
includes.remove('z')

data, full_parameter_dict = hddm.simulators.hddm_dataset_generators.simulator_h_c(n_subjects = 5,
                                                                                  n_trials_per_subject = n_samples,
                                                                                  model = model,
                                                                                  p_outlier = 0.00,
                                                                                  conditions = None,
                                                                                  depends_on = None,
                                                                                  regression_models = None,
                                                                                  regression_covariates = None,
                                                                                  group_only_regressors = False,
                                                                                  group_only = None,
                                                                                  fixed_at_default = ['z'])

data
rt response subj_idx v0 v1 v2 v3 a z t theta
0 1.723548 1.0 0 0.973732 2.066200 0.541262 1.266078 1.988639 0.5 1.559548 0.656266
1 1.784549 2.0 0 0.973732 2.066200 0.541262 1.266078 1.988639 0.5 1.559548 0.656266
2 2.076545 3.0 0 0.973732 2.066200 0.541262 1.266078 1.988639 0.5 1.559548 0.656266
3 1.811549 3.0 0 0.973732 2.066200 0.541262 1.266078 1.988639 0.5 1.559548 0.656266
4 1.748549 3.0 0 0.973732 2.066200 0.541262 1.266078 1.988639 0.5 1.559548 0.656266
... ... ... ... ... ... ... ... ... ... ... ...
4995 1.931315 1.0 4 0.622314 1.956002 0.706355 1.264215 2.106691 0.5 1.569316 0.661371
4996 1.851316 1.0 4 0.622314 1.956002 0.706355 1.264215 2.106691 0.5 1.569316 0.661371
4997 1.910315 3.0 4 0.622314 1.956002 0.706355 1.264215 2.106691 0.5 1.569316 0.661371
4998 1.751316 1.0 4 0.622314 1.956002 0.706355 1.264215 2.106691 0.5 1.569316 0.661371
4999 1.908315 0.0 4 0.622314 1.956002 0.706355 1.264215 2.106691 0.5 1.569316 0.661371

5000 rows × 11 columns

# Define the HDDM model
hddmnn_model = hddm.HDDMnn(data,
                           informative = False,
                           include = includes,
                           p_outlier = 0.0,
                           model = model)

# Sample
hddmnn_model.sample(nmcmc,
                    burn = 100)

[-----------------100%-----------------] 1001 of 1000 complete in 475.6 sec

<pymc.MCMC.MCMC at 0x106755050>

caterpillar plot

# Caterpillar Plot: (Parameters recovered ok?)
hddm.plotting.plot_caterpillar(hddm_model = hddmnn_model,
                               ground_truth_parameter_dict = full_parameter_dict,
                               drop_sd = False,
                               figsize = (8, 6))
plt.show()
_images/lan_visualizations_30_0.png

posterior_predictive_plot

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (8, 6),
                                        value_range = np.arange(0, 4, 0.02),
                                        plot_func = hddm.plotting._plot_func_posterior_node_from_sim,
                                        parameter_recovery_mode = True,
                                        **{'alpha': 0.01,
                                           'ylim': 3,
                                           'bin_size': 0.1,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_rts': False,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.3, 'wspace': 0.3}})
_images/lan_visualizations_32_0.png

model_cartoon_plot

WARNING: The plot below should not be taken as representative for a particular model fit. The chain may need to be run for much longer than the number of samples allotted in this tutorial script.

hddm.plotting.plot_posterior_predictive(model = hddmnn_model,
                                        columns = 2, # groupby = ['subj_idx'],
                                        figsize = (12, 8),
                                        value_range = np.arange(0, 3, 0.1),
                                        plot_func = hddm.plotting._plot_func_model_n,
                                        parameter_recovery_mode = True,
                                        **{'add_legend': False,
                                           'alpha': 0.01,
                                           'ylim': 6.0,
                                           'bin_size': 0.025,
                                           'add_posterior_mean_model': True,
                                           'add_posterior_mean_rts': True,
                                           'add_posterior_uncertainty_model': True,
                                           'add_posterior_uncertainty_rts': False,
                                           'samples': 200,
                                           'legend_fontsize': 7,
                                           'legend_loc': 'upper left',
                                           'linewidth_histogram': 1.0,
                                           'subplots_adjust': {'top': 0.9, 'hspace': 0.35, 'wspace': 0.3}})
plt.show()
_images/lan_visualizations_35_0.png