Robust unified Granger causality analysis: a normalized maximum likelihood form

Brain Informatics

Table. 2 Comparison between uGCA methods and conventional GCA under different data length

	uGCA-TP (%)	uGCA-MIX (%)	uGCA-NML (%)	GCA (\(\alpha =0.05\)) (%)	GCA (\(\alpha =0.01\)) (%)
\(L=150\)
TPR	89.933	83.433	86.178	93.810	86.200
TNR	99.048	99.710	99.624	95.743	98.705
Ground-truth rate	31	17.5	22	24.4	18.5
\(L=200\)
TPR	95.378	91.411	92.967	97.200	93.289
TNR	99.181	99.748	99.705	95.610	98.610
Ground-truth rate	56.4	41.1	48.2	29.3	40.8
\(L=300\)
TPR	99.567	98.022	99.133	99.620	98.878
TNR	99.433	99.857	99.820	98.878	98.890
Ground-truth rate	85.5	83.0	88.9	38.3	72.1
\(L=500\)
TPR	99.989	99.033	99.989	100	100
TNR	99.324	99.824	99.786	95.186	98.571
Ground-truth rate	86.3	91.3	95.6	34.2	72.9

The data length ranges from 150 to 500, the results (\(L=1000\)) were present in Table 1, and L is the data length
Bold values indicate more indicative of the method's performance