Table 2

Simulation results on estimation.




Model Parameter Estimates
 Reg of h on Math




setting
 true # z
 used # z
 n
 β
 ρ
 Intercept
 Slope
 R2
1
 5
 5
 100
 1.10
 71.50a(estimated)
 -0.06
 1.06
 0.82




1.14
 1.00 (fixed)
 -0.28
 1.48
 0.79




1.08
 20.00 (fixed)
 -0.08
 1.15
 0.84
2
 5
 5
 200
 0.99
 90.03 (estimated)
 0.01
 1.04
 0.87




1.05
 1.00 (fixed)
 -0.01
 1.13
 0.84




0.96
 20.00 (fixed)
 -0.00
 1.07
 0.87
3
 5
 5
 300
 0.98
 111.76 (estimated)
 -0.01
 1.04
 0.90




1.03
 1.00 (fixed)
 -0.02
 1.10
 0.87




0.97
 20.00 (fixed)
 -0.01
 1.06
 0.90

This table shows the simulation results of estimated regression coefficients βand the nonparametric function h(·) in model logit(π) = xβ+ h(z) for binary outcomes based on 300 runs. True β = 1. In the table, a is the average of the estimated Math from 300 simulations.

Liu et al. BMC Bioinformatics 2008 9:292   doi:10.1186/1471-2105-9-292