Skip to main content

Table 2 The average of performance of various model selection algorithms for the four simulation scenarios using four model selection methods in term of the fraction of correctly estimated link/non-links, i.e. false positives (FP), false negatives (FN), false discoveries (FD) and false non-discoveries (FnD), as well as the F1 = (2 - 2FN)/(2 - FN + FP) score, which measures the overall average accuracy of recall and precision. The best scores are indicated by bold font.

From: Inferring slowly-changing dynamic gene-regulatory networks

p

 

F P ¯

F N ¯

F D ¯

F n D ¯

F1score

20

AICc

0.0092

0.0811

0.2000

0.0031

0.9532

 

BIC

0.0363

0.0139

0.4873

0.0005

0.9751

 

AIC

0.0698

0.0069

0.6470

0.0003

0.9628

40

AICc

0.0057

0.0447

0.2899

0.0006

0.9743

 

BIC

0.0088

0.0321

0.3826

0.0005

0.9793

 

AIC

0.0437

0.0041

0.7514

0.0001

0.9766

60

AICc

0.0016

0.4585

0.2730

0.0036

0.7018

 

BIC

0.0016

0.4585

0.2730

0.0036

0.7018

 

AIC

0.0288

0.1452

0.8088

0.0012

0.9076

80

AICc

0.0091

0.1034

0.1680

0.0052

0.9410

 

BIC

0.0396

0.0517

0.4527

0.0027

0.9541

 

AIC

0.0670

0.0000

0.5704

0.0000

0.9675