Application of Seemingly Unrelated Regression in Medical Data with Intermittently Observed Time-Dependent Covariates
Table 2
Simulation results (summary of estimated parameters, bias, and MSE) for complete observation of time-dependent covariates with different sample sizes (, and 240).
Parameter (true value)
Method
Sample size = 20
Sample size = 60
Sample size = 120
Sample size = 240
Bias
MSE
Bias
MSE
Bias
MSE
Bias
MSE
REML
−1.01
.12
7.24
−1.14
−.009
1.89
−1.12
.01
.828
−1.14
−.006
.42
(−1.13)
ITSUR
−1.003
.13
7.37
−1.14
−.009
1.89
−1.12
.01
.829
−1.14
−.006
.42
REML
.18
−.04
7.33
.14
−.09
1.90
.21
−.02
.864
.25
.02
.42
(.23)
ITSUR
.18
−.05
7.40
.14
−.09
1.89
.21
−.02
.863
.25
.02
.42
REML
.03
.04
3.32
−.01
−.004
.82
−.02
−.009
.39
−.02
−.009
.18
(−.01)
ITSUR
.02
.03
3.39
−.009
.0004
.83
−.02
−.008
.39
−.02
−.009
.18
REML
−1.11
.006
2.25
−1.12
−.003
.57
−1.08
.04
.26
−1.10
.02
.13
(−1.12)
ITSUR
−1.11
.01
2.36
−1.12
−.005
.58
−1.09
.03
.27
−1.10
.02
.13
*Note: REML represents the estimated parameters from MRM, and ITSUR represents that of the SUR model.