.

Author: smilesun

causalspyne/utils_cov_precision.py

@@ -0,0 +1,8 @@
+import numpy as np
+
+def estimate_spectrum_radius(arr_df):
+    mat_corr = np.corrcoef(arr_df.T)
+    mat_cov = np.cov(arr_df.T)
+    mat_precision = np.linalg.inv(mat_cov)
+    arr_eig_values, eigvec = np.linalg.eig(mat_precision)
+    return np.max(arr_eig_values)

examples/example.py

@@ -1,9 +1,13 @@
 from causalspyne import gen_partially_observed
+from causalspyne.utils_cov_precision import estimate_spectrum_radius
 
 
-gen_partially_observed(size_micro_node_dag=3,
-                       num_macro_nodes=3,
-                       degree=2,  # average vertex/node degree
-                       list_confounder2hide=[0.1, 1.0], # choie of confounder to hide: percentile or index of all toplogically sorted confounders
-                       num_sample=200,
-                       graphviz=True)
+subview = gen_partially_observed(size_micro_node_dag=3,
+                                 num_macro_nodes=3,
+                                 degree=2,
+                                 list_confounder2hide=[0],
+                                 num_sample=200,
+                                 graphviz=True)
+
+radius = estimate_spectrum_radius(subview.data)
+print(radius)