Results Dictionary

For a single target analysis, the analyse_single_target method returns a results dictionary which looks like this:

In [1]: res
Out[1]:
{'current_value': (3, 5),  # process id and sample of the current value
 'max_lag_sources': 5,  # lags specified by the user
 'max_lag_target': 5,
 'min_lag_sources': 1,
 'omnibus_pval': 0.047619047619047616,  # p-value of the omnibus test
 'omnibus_sign': True,  # significance of the omnibus test
 'omnibus_te': 0.37424613847591426,  # estimated omnibus TE
 'options': {'cmi_calc_name': 'jidt_kraskov',  # analysis opts provided by the user
  'n_perm_max_seq': 21,
  'n_perm_max_stat': 21,
  'n_perm_min_stat': 21,
  'n_perm_omnibus': 21},
 'selected_vars_pval': array([ 0.04761905]),  # indiv. p-values for selected source variables
 'selected_vars_te': array([ 0.37424614]),  # indiv. TE values for selected source variables
 'selected_vars_full': [(3, 1), (3, 2), (3, 3), (0, 2)],  # all selected variables, full set Z
 'selected_vars_sources': [(0, 2)],  # selected variables from the sources' past, Z_X
 'selected_vars_target': [(3, 1), (3, 2), (3, 3)],  # selected variables from the target's past, Z_Y
 'sources_tested': [0, 1, 2, 4],  # potential source processes entering the analysis
 'target': 3,  # target process
 'tau_sources': 1,  # tau used for defining the source candidate set
 'tau_target': 1}  # tau used for defining the target candidate set

For multiple targets analysis, the analyse_network method returns a results dictionary containing one entry for each analysed target. Each entry contains a results dictionary for a single target, which looks like the dictionary above.

In [1]: res = network_analysis.analyse_network(data=dat)
In [2]: res.keys()
Out[2]: dict_keys([0, 1, 2, 3, 4, 'fdr'])

Additionally, the dictionary contains an entry 'fdr', which contains the results dictionary after FDR-correction over targets (Benjamini, 1995, J Royal Stat Soc B, 57(1)).