The Use of Mixed Models for the Analysis of Mediated Data with Time-Dependent Predictors
Table 3
Performance of unrestricted structural equation model (USEM) and unrestricted and time interaction linear mixed models (UMM and TIMM, resp.) when true underlying effects are small for early time points and small to moderate for late time points.
Simulated data scenarios
USEM
TIMM
UMM
Time point
Effect size
Sample size
Bias (%)
Coverage probability (%)
Power (%)
Bias (%)
Coverage probability (%)
Power (%)
Bias (%)
Coverage probability (%)
Power (%)
0.05
350
16
95
8
93
6
16
96
7
0.05
350
9.4
94
8
51
94
19
8.6
94
8
0.05
350
7.4
94
9
153
75
56
4.9
95
8
0.25
350
−0.6
94
54
−29
81
76
−0.09
94
54
0.25
350
−0.4
94
48
−8.9
94
73
−0.5
93
48
0.25
350
−0.08
93
40
11
93
66
−1.0
94
38
0.05*
350
−4.0
94
8
−71
94
5
−6.3
95
7
0.05*
350
−9.6
96
7
38
94
15
−9.8
95
6
0.05*
350
−1.4
95
8
146
78
49
1.3
95
7
0.25*
350
2.9
95
55
−29
81
70
2.6
94
54
0.25*
350
2.2
95
49
−7.4
94
71
1.6
94
48
0.25*
350
0.4
95
36
14
94
66
−0.5
95
36
0.05**
350
10
95
7
−60
94
6
8.7
95
7
0.05**
350
93
7
44
95
18
93
7
0.05**
350
11
94
8
148
75
56
12
94
9
0.25**
350
−0.4
95
58
−29
81
79
−0.3
95
55
0.25**
350
0.0
96
50
−8.6
95
77
0.2
96
48
0.25**
350
−0.2
96
37
12
95
70
−2.6
96
36
0.05
100
94
7
−26
94
6
93
6
0.05
100
8.8
94
6
7.2
93
6
36
94
9
0.05
100
10
96
5
14
95
5
141
90
20
0.25
100
−2.0
94
20
−2.2
93
20
90
30
0.25
100
−3.8
94
17
−3.2
94
17
−9.8
95
29
0.25
100
−2.5
94
16
−2.2
94
16
11
95
26
0.05
200
8.2
95
7
−60
95
5
7.7
95
7
0.05
200
1.2
95
7
45
95
13
0.4
95
6
0.05
200
−7.8
95
7
150
83
34
−9.2
95
6
0.25
200
4.0
96
38
−29
97
52
4.1
95
36
0.25
200
−3.0
95
28
−8.1
95
50
−2.6
95
28
0.25
200
−0.7
94
24
13
95
45
−1.3
94
25
0.05
400
41
95
8
−65
94
5
2.7
95
7
0.05
400
−3.2
94
7
41
94
19
−4.7
97
6
0.05
400
−0.2
95
9
146
74
58
−2.3
96
7
0.25
400
−1.3
95
60
−30
76
80
−1.7
95
58
0.25
400
1.2
95
54
−8.6
95
79
1.0
94
53
0.25
400
−0.5
96
43
12
95
73
−1.9
96
40
0.05
400
−6
95
7
52
94
14
−7
95
7
0.05
400
4
97
6
−33
95
8
3
97
7
0.05
400
2
95
8
−119
82
5
5
95
7
−0.11
400
0.4
95
17
−52
83
14
0.9
95
17
−0.11
400
2.5
95
17
−13
95
23
4
95
17
−0.11
400
1
95
13
26
94
26
2
95
13
Based on 1000 simulated datasets. Results are from simulated data with total effects equally distributed between direct and indirect effects, except where indicated.
*Total effect is primarily direct.
**Total effect is primarily indirect.