Skip to main content

Table 3 Simulation study results for the other six models

From: Two-way learning with one-way supervision for gene expression data

 

G=2,…,10

G=g known

Model

Average G

Most chosen model

Average ARI

Average ARI

g known = 2

    

UUU

2

UUC

1.0 (0.0)

NA

UUC

2

UUC

1.0 (0.003)

NA

UCU

2

UCC

0.960 (0.028)

NA

UCC

2

UCC

0.961 (0.025)

NA

CCU

2.3 (0.7)

CCU

0.971 (0.098)

1.0 (0.0)

CCC

2.4 (1.2)

CCU

0.959 (0.113)

1.0 (0.0)

g known = 3

    

UUU

3

UUC

1.0 (0.0)

NA

UUC

3

UUC

1.0 (0.0)

NA

UCU

3

UCC

1.0 (0.0)

NA

UCC

3

UCC

1.0 (0.0)

NA

CCU

4.0 (1.3)

CCU

0.936 (0.095)

1.0 (0.0)

CCC

4.3 (1.4)

CCU

0.915 (0.107)

1.0 (0.0)

g known = 4

    

UUU

4

UUC

1.0 (0.0)

NA

UUC

4

UUC

1.0 (0.0)

NA

UCU

4

UCC

1.0 (0.0)

NA

UCC

4

UCC

1.0 (0.0)

NA

CCU

5.1 (1.3)

CCU

0.958 (0.057)

1.0 (0.055)

CCC

5.1 (1.1)

CCU

0.960 (0.052)

1.0 (0.029)

  1. Average ARI, most chosen model, and the average number of observation clusters selected for the remaining six OSGaBI models using simulated data with medium variance and good cluster separation when fitting G=2,…,10 observation clusters using 100 data sets and 20 random starts. The last column presents the ARI when fixing G=g known, where g known represents the number of observation clusters the data was generated from, for the CCU and CCC models. Values in brackets represent the respective standard deviation