Skip to main content

Table 8 Selected logistic regression models (LRM).

From: Neural networks for modeling gene-gene interactions in association studies

  

LRM with design variables

  

Statistical model (# parameters)

Two-locus disease model

Risk scenario

NM (1)

SiA (3)

SiB (3)

ME (5)

FM (9)

ADD

low

  

1

39

60

100

 

high

   

7

93

100

HET

low

   

55

45

100

 

high

   

10

90

100

MULT

low

   

90

10

100

 

high

   

88

12

100

EPI RR

low

  

6

 

94

100

 

high

    

100

100

EPI DD

low

   

3

97

100

 

high

    

100

100

EPI RD

low

 

61

 

19

20

100

 

high

 

57

 

14

29

100

  

LRM without design variables

  

Statistical model (# parameters)

  

NM (1)

SiA (2)

SiB (2)

ME (3)

FM (4)

ADD

low

  

1

27

72

100

 

high

   

4

96

100

HET

low

   

30

70

100

 

high

   

6

94

100

MULT

low

   

72

28

100

 

high

   

54

46

100

EPI RD

low

7

6

9

3

75

100

 

high

    

100

100

EPI DD

low

   

2

98

100

 

high

    

100

100

EPI RD

low

 

60

 

19

21

100

 

high

 

38

 

23

39

100

  1. In the upper part of the table, the two-locus disease model (ADD, HET) agrees with the statistical model when a statistical model of independence (NM, SiA, SiB, ME) is selected. In the lower part of the table, the two-locus disease model representing biological interaction (MULT, EPI RR, EPI DD, EPI RD) agrees with the statistical model when the full model (FM) is selected. Bold numbers mark the mode of the selected models in the low and high risk scenario.