Skip to main content

Table 2 Summary of results

From: Component retention in principal component analysis with application to cDNA microarray data

Column No.

1

2

3

4

5

6

7

8

Data Set

Y + N0

Y + N.25

alpha

cdc15

elutriation

fibro

sporula

tumor

Broken Stick, (BS)

1

1

2

4

3

2

1

3

Modified BS

1

1

2

4

3

2

1

3

Velicer's MAP

2/8

2

3

5

3

3

2

8/10

Kaiser-Guttman, (KG)

2

2

3

4

3

3

2

9

Jolliffe's KG

7

7

4

5

4

3

3

14

LEV Diagram

6/8

6/8

4/5

5

4/5

4/6

3

12/21

Parallel Analysis

1

1

5

4

3

3

2

8

Scree Test

8

8

5

5

4

6

4

7

Info Dimension

5.9

7.3

11.1

7.2

6.4

3.0

2.6

17.3

Gleason-Staelin Stat

.525

.45

.34

.37

.38

.54

.58

.37

Normalized Entropy

.917

.941

.779

.726

.706

.438

.493

.696

80% of Var.

4

4

9

4

5

3

3

19

90% of Var.

7

7

14

7

9

5

3

33

Bartlett's Test

15

15

22

15

14

12

7

60

  1. The table contains the results of twelve stopping rules along with two measures of data information content for six cDNA microarray data sets. We recommend looking for a consensus among the rules given in the upper portion of the table, while avoiding rules based on cumulative percent of variation explained or Barlett's test. Synthetic data sets are summarized in columns 1 (no noise) and 2 (Gaussian noise, μ = 0 and σ = 0.25). The matrix sizes for columns 1 through 8 are : (6000 × 15), (6000 × 15), (4579 × 22), (799 × 15), (5981 × 14), (517 × 12), (6118 × 7), and (1375 × 60).