Plotting¶

CDANs ships two plotting functions, both wrappers around tigramite.plotting:

plot_process_graph — one node per variable, with curved arrows for lagged edges and straight arrows for contemporaneous ones. The surrogate C is materialized as an extra node connected to changing modules.
plot_time_series_graph — time-unrolled, one node per (variable, time-step). The surrogate C is drawn as an extra row at the top (by default), with arrows down to changing modules at every time-step. Changing modules are also highlighted in orange.

Install¶

Plotting is an optional extra:

pip install "cdans[viz]"

This brings in matplotlib and tigramite. The core library does not require either at runtime — only when you call a plotting function.

Quick example¶

from cdans import CDANs, plot_process_graph, plot_time_series_graph
from cdans.utils import generate_synthetic_cdans

dataset = generate_synthetic_cdans(n_vars=5, n_samples=400, tau_max=2, seed=0)
result = CDANs(tau_max=2, ci_test="kci").fit(dataset.data)

# Process graph (aggregate, with surrogate C drawn explicitly)
plot_process_graph(
    result.graph,
    var_names=["HR", "BP", "SpO2", "RR", "Temp"],
    save_path="process.png",
)

# Time-series graph (time-unrolled, changing modules in orange)
plot_time_series_graph(
    result.graph,
    var_names=["HR", "BP", "SpO2", "RR", "Temp"],
    save_path="time_series.png",
)

Customizing the plot¶

Both functions accept the same **kwargs as their tigramite counterparts. A few that are worth knowing:

Argument	Purpose
`figsize=(w, h)`	Matplotlib figure size in inches
`arrow_linewidth=4`	Edge thickness
`node_size=0.4`	Node radius (process graph)
`label_fontsize=12`	Variable-name font size
`fig_ax=(fig, ax)`	Draw into an existing matplotlib axis
`save_path="out.png"`	Save instead of just returning the figure
`show_surrogate=False`	Hide the `C` node entirely
`surrogate_position="bottom"`	Move the `C` row from top to bottom (time-series graph only)

For the process graph specifically, show_surrogate=False hides the C node if you'd rather see only the data variables.

Output format¶

Both functions return a (fig, ax) matplotlib pair so you can post-process or embed in a larger figure. The save_path keyword is a convenience that calls fig.savefig(save_path, bbox_inches="tight", dpi=150).

What the link types mean¶

In tigramite's plotting convention:

Solid straight arrow between two nodes: directed contemporaneous edge (e.g. X1 -> X2).
Empty-circle endpoints (small unfilled circles instead of arrowheads): undirected contemporaneous edge — the algorithm couldn't determine a direction.
Curved arrow with a number label: lagged edge; the number is the lag.
Edges from C: the surrogate variable's effect on changing modules. Visible in both views by default; turn off with show_surrogate=False.