An Eigenvalue test for spatial principal component analysis

BMC Bioinformatics

Table 3 Results of the spca_randtest with 10,000 permutations on the human mtDNA dataset (Montano et al., [14])

Spatial patterns	Observed p-value	Decreasing Positive Eigenvalues	Observed p-value	Bonferroni corrected significant level
Global pattern	0.0058	3.4e-2	0.0105	0.05
Local pattern	0.8826	8.5e-3	0.0137	0.025
		4.1e3	0.0136	0.016
		1.6e-3	0.506	0.0125

The simulated distribution of the f _i ⁺ and f _i ⁻ statistics are compared to the f _i ⁺ and f _i ⁻ statistics observed for the original dataset. A significant global pattern (or significant f _i ⁺ observed statistics) is found with the spca_randtest (p-value <0.01). Thus, each positive eigenvalue is compared with its simulated distribution and assigned to be significant if its observed p-value is lower than the corrected Bonferroni p-value, with starting threshold of 0.05. Significant observed p-values as compared with Bonferroni corrected p-values are highlighted in bold

ISSN: 1471-2105