Importance of data structure in comparing two dimension reduction methods for classification of microarray gene expression data

Table 2 Proportion of well-classified patients according to ratio, which reflects eccentricity

	α = π/2	α = π/3	α = π/4	α = π/6	α = 0
ratio = 10
PLS+DA	0.82(0.05)//2	0.81(0.03)//2	0.76(0.03)//2	0.71(0.04)//2	0.59(0.04)//2
PCA+DA	0.85(0.05)//3	0.81(0.04)//3	0.77(0.05)//3	0.73(0.05)//3	0.59(0.04)//3
BGA	0.76(0.05)	0.75(0.04)	0.66(0.03)	0.67(0.04)	0.58(0.04)
ratio = 2
PLS+DA	0.68(0.05)//1	0.65(0.04)//1	0.65(0.05)//1	0.65(0.05)//2	0.63(0.05)//1
PCA+DA	0.69(0.05)//3	0.65(0.04)//3	0.67(0.04)//2	0.65(0.04)//2	0.63(0.04)//2
BGA	0.67(0.06)	0.62(0.04)	0.64(0.05)	0.65(.04)	0.62(0.05)
ratio = 1
PLS+DA	0.60(0.05)//2	0.62(0.05)//1	0.64(0.05)//2	0.62(0.05)//1	0.61(0.05)//1
PCA+DA	0.63(0.04)//3	0.63(0.04)//2	0.63(0.05)//2	0.64(0.05)//2	0.63(0.05)//2
BGA	0.61(0.05)	0.62(0.05)	0.61(0.05)	0.61(0.05)	0.60(0.05)

Mean (Standard deviation)//Median of the optimal number of components over over fifty datasets simulated with a distance between the barycenters dist = 2. PLS: Partial Least Squares – PCA: Principal Components Analysis – DA: Discriminant Analysis.

ISSN: 1471-2105