Table 7B.

Post hoc comparisons.

Task	Tractography algorithm	(I) Dimension	(J) Dimension	Mean difference (I-J)	Sig.	95% confidence interval
						Lower bound	Upper bound
AD vs. MCI	Probtrackx	5	10	0.11181	0.016	0.0111	0.2126
			25	0.10728	0.026	0.0065	0.2080
			30	0.12289	0.005	0.0221	0.2236
			35	0.13449	0.001	0.0337	0.2352
MCI vs. NC	Tensor-SL	5	10	-0.09339	0.045	-0.1857	-0.0011
			35	-0.10498	0.012	-0.1973	-0.0127

The “Sig.” column shows the SPSS adjusted p-value; only values 0.05 are treated as significant. Only comparisons that passed Bonferroni correction are listed here. Even for the situations listed in this table, there is no consistent trend in the comparison between higher and lower dimensions. For example in the task of classifying AD vs. MCI, using Probtrackx, the lower dimension setting (5) has better performance than higher dimensional settings (10, 25, 30, and 35), as mean differences are all positive; however, for the task MCI vs. NC, the trend is opposite when using Tensor-SL.