Skip to main content

Table 4 Realized βa, mean bias, mean standard error (SE), coverage probability and percentage of the effects that had a p-value < 0.05 in the simulation where haplotypes 001 and 101 had a modeled parameter estimate of 1.10. Each replicate contained 2000 individuals.

From: Estimating effects of rare haplotypes on failure time using a penalized Cox proportional hazards regression model

No penalty

Haplotype

Frequency

Realizeda

Mean bias

Mean SE

Coverage

P < 0.05b

001

0.050

1.094

-0.005

0.08

0.93

1.00

101

0.003

1.155

-0.037

0.38

0.93

0.77

010

0.020

0.016

-0.007

0.14

0.98

0.02

011

0.005

0.023

0.036

0.30

0.98

0.02

100

0.020

0.006

0.015

0.14

1.00

0.00

110

0.002

0.084

0.075

1.14

0.95

0.05

111

0.280

-0.007

-0.007

0.03

0.82

0.18

Ridge penalty

Haplotype

Frequency

Realizeda

Mean bias

Mean SE

Coverage

P < 0.05b

001

0.050

1.094

-0.009

0.07

0.91

1.00

101

0.003

1.155

-0.085

0.61

0.91

0.62

010

0.020

0.016

-0.007

0.013

0.96

0.04

011

0.005

0.023

0.046

0.23

0.96

0.04

100

0.020

0.006

0.014

0.12

0.98

0.02

110

0.002

0.084

0.067

0.40

0.66

0.34

111

0.280

-0.007

-0.008

0.03

0.81

0.19

Difference penalty

Haplotype

Frequency

Realizeda

Mean bias

Mean SE

Coverage

P < 0.05b

0 0 1

0.050

1.094

-0.011

0.07

0.88

1.00

1 0 1

0.003

1.155

-0.066

0.54

0.88

0.71

0 1 0

0.020

0.016

0.004

0.12

0.96

0.04

0 1 1

0.005

0.023

0.071

0.22

0.90

0.10

1 0 0

0.020

0.006

0.017

0.12

0.98

0.02

1 1 0

0.002

0.084

0.069

0.37

0.57

0.43

1 1 1

0.280

-0.007

-0.009

0.03

0.82

0.18

  1. a) "Realized" denotes to the median estimate from the simulated data, where the haplotypes were known, and could be used in the Cox model directly; b) For haplotypes 001 and 101 these percentages reflect power, whereas the percentages for the other five haplotypes reflect type I error probabilities.