Hi all,
I have the following data frame named "one" in which there is a
grouping factor method with three levels. The varibility is very
different among the three groups seen from the plot. I am trying to
compare three models as follows. I have the following two questions:
1. As you can see the third model does not seem to work. I am not
sure whether it is a convergent issue.
2. The first model doesnot take into account the variance
heterogeneity issue while the second one does. However, when I compare
the residual plot between mod1 and mod2, there is not much difference.
The residual for the second model still has much larger variance for
methods 2 and 3 than the method 1.
Thanks very much!
Hanna
method individual time res
1 3 12 0 101.40000
2 3 12 3 101.50000
3 3 12 6 101.50000
4 3 12 9 101.30000
5 3 12 12 100.70000
6 3 12 15 101.00000
7 3 12 18 101.50000
14 3 10 0 101.30000
15 3 10 3 101.20000
16 3 10 6 101.50000
17 3 10 9 100.70000
18 3 10 12 101.50000
19 3 10 15 101.30000
20 3 10 18 101.30000
27 3 11 0 100.70000
28 3 11 3 101.10000
29 3 11 6 101.90000
30 3 11 9 100.80000
31 3 11 12 99.80000
32 3 11 15 100.60000
33 3 11 18 100.60000
40 3 1 0 97.50000
41 3 1 3 97.40000
42 3 1 6 97.70000
43 3 1 9 97.40000
44 3 1 12 97.30000
45 3 1 15 96.70000
46 3 1 18 96.60000
54 3 3 0 98.10000
55 3 3 3 98.50000
56 3 3 6 97.90000
57 3 3 9 97.90000
58 3 3 12 97.70000
59 3 3 15 98.00000
60 3 3 18 98.50000
67 3 6 0 100.20000
68 3 6 3 99.60000
69 3 6 6 99.90000
70 3 6 9 99.90000
71 3 6 12 100.30000
77 3 2 0 98.90000
78 3 2 3 98.90000
79 3 2 6 98.70000
80 3 2 9 98.90000
81 3 2 12 98.80000
82 3 2 15 97.80000
83 3 2 18 98.90000
90 3 4 0 100.20000
91 3 4 3 99.80000
92 3 4 6 99.50000
93 3 4 9 100.40000
96 3 5 0 100.70000
97 3 5 3 100.30000
98 3 5 6 100.70000
99 3 5 9 100.50000
102 3 7 0 100.90000
105 3 8 0 99.30000
108 3 9 0 100.20000
111 3 13 0 101.00000
114 3 14 0 100.80000
117 3 15 0 100.40000
8 2 12 0 108.00000
9 2 12 12 97.00000
21 2 10 0 112.00000
22 2 10 12 93.00000
34 2 11 0 98.00000
35 2 11 12 96.00000
47 2 1 0 94.00000
48 2 1 12 103.00000
49 2 1 18 103.00000
61 2 3 0 87.00000
62 2 3 12 105.00000
72 2 6 0 119.00000
73 2 6 3 105.00000
74 2 6 12 91.00000
84 2 2 0 105.00000
85 2 2 12 112.00000
95 2 4 0 96.00000
101 2 5 0 113.00000
104 2 7 0 106.00000
107 2 8 0 71.00000
110 2 9 0 95.00000
113 2 13 0 88.00000
116 2 14 0 86.00000
119 2 15 0 81.00000
86 1 12 0 105.90300
94 1 12 12 99.82400
10 1 12 15 91.26400
11 1 12 18 72.15000
191 1 10 0 91.14300
201 1 10 12 100.36800
211 1 10 15 104.79600
221 1 10 18 102.58200
301 1 11 0 78.32400
311 1 11 12 88.20900
321 1 11 15 95.52600
331 1 11 18 106.87200
411 1 1 0 99.04500
421 1 1 12 82.83600
431 1 1 15 116.51200
441 1 1 18 89.05600
52 1 3 0 97.81800
53 1 3 12 89.81500
541 1 3 15 82.11000
551 1 3 18 79.83400
611 1 6 0 89.06000
621 1 6 12 102.56500
701 1 2 0 112.32000
711 1 2 12 104.40000
721 1 2 15 90.06800
731 1 2 18 107.28000
781 1 4 0 125.92500
831 1 5 0 95.16000
851 1 7 0 111.28981
87 1 8 0 102.22482
89 1 9 0 91.61610
911 1 13 0 111.18053
931 1 14 0 91.70376
951 1 15 0 98.04994
library(nlme)
mod1 <- lme(fixed= res ~ method*time, random=~ 1+ time | individual,
data=one)
summary(mod1)
anova(mod1)
mod2 <- lme(fixed= res ~ method*time, random=~ 1+ time | individual,
data=one, weights= varIdent(form=~1|method))
summary(mod2)
anova(mod2)
mod3 <- lme(fixed= res ~ method*time, random=~ 0+ method+ time |
individual, data=one, weights= varIdent(form=~1|assay))
summary(mod3)
anova(mod3)
par(mfrow=c(1,2))
plot(resid(mod1))
plot(resid(mod2))