[R-sig-ME] Trouble Replicating Unstructured Mixed Procedure in R
Thompson,Paul
Paul.Thompson at SanfordHealth.org
Wed Jan 25 03:48:25 CET 2012
In the CS model, the F values for Gender and Gender*age are really close, but age is quite discrepant. That seems problematic.
-----Original Message-----
From: r-sig-mixed-models-bounces at r-project.org [mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf Of Charles Determan Jr
Sent: Tuesday, January 24, 2012 8:32 PM
To: r-sig-mixed-models at r-project.org
Subject: [R-sig-ME] Trouble Replicating Unstructured Mixed Procedure in R
Greetings,
I have been working on R for some time now and I have begun the endeavor of
trying to replicate some SAS code in R. I have scoured the forums but
haven't been able to find an answer. I hope one of you could be so kind as
to enlighten me.
I am attempting to replicate a repeated measures experiment using some
standard data. I have posted the SAS code and output directly from a
publication as well as my attempts in R to replicate it. My main issue
comes with the 'unstructured' component.
The 'dental' dataset from 'mixedQF' package,
equivalent to formixed data in SAS
distance age Subject Sex
1 26.0 8 M01 Male
2 25.0 10 M01 Male
3 29.0 12 M01 Male
4 31.0 14 M01 Male
5 21.5 8 M02 Male
6 22.5 10 M02 Male
7 23.0 12 M02 Male
8 26.5 14 M02 Male
9 23.0 8 M03 Male
10 22.5 10 M03 Male
11 24.0 12 M03 Male
12 27.5 14 M03 Male
13 25.5 8 M04 Male
14 27.5 10 M04 Male
15 26.5 12 M04 Male
16 27.0 14 M04 Male
17 20.0 8 M05 Male
18 23.5 10 M05 Male
19 22.5 12 M05 Male
20 26.0 14 M05 Male
21 24.5 8 M06 Male
22 25.5 10 M06 Male
23 27.0 12 M06 Male
24 28.5 14 M06 Male
25 22.0 8 M07 Male
26 22.0 10 M07 Male
27 24.5 12 M07 Male
28 26.5 14 M07 Male
29 24.0 8 M08 Male
30 21.5 10 M08 Male
31 24.5 12 M08 Male
32 25.5 14 M08 Male
33 23.0 8 M09 Male
34 20.5 10 M09 Male
35 31.0 12 M09 Male
36 26.0 14 M09 Male
37 27.5 8 M10 Male
38 28.0 10 M10 Male
39 31.0 12 M10 Male
40 31.5 14 M10 Male
41 23.0 8 M11 Male
42 23.0 10 M11 Male
43 23.5 12 M11 Male
44 25.0 14 M11 Male
45 21.5 8 M12 Male
46 23.5 10 M12 Male
47 24.0 12 M12 Male
48 28.0 14 M12 Male
49 17.0 8 M13 Male
50 24.5 10 M13 Male
51 26.0 12 M13 Male
52 29.5 14 M13 Male
53 22.5 8 M14 Male
54 25.5 10 M14 Male
55 25.5 12 M14 Male
56 26.0 14 M14 Male
57 23.0 8 M15 Male
58 24.5 10 M15 Male
59 26.0 12 M15 Male
60 30.0 14 M15 Male
61 22.0 8 M16 Male
62 21.5 10 M16 Male
63 23.5 12 M16 Male
64 25.0 14 M16 Male
65 21.0 8 F01 Female
66 20.0 10 F01 Female
67 21.5 12 F01 Female
68 23.0 14 F01 Female
69 21.0 8 F02 Female
70 21.5 10 F02 Female
71 24.0 12 F02 Female
72 25.5 14 F02 Female
73 20.5 8 F03 Female
74 24.0 10 F03 Female
75 24.5 12 F03 Female
76 26.0 14 F03 Female
77 23.5 8 F04 Female
78 24.5 10 F04 Female
79 25.0 12 F04 Female
80 26.5 14 F04 Female
81 21.5 8 F05 Female
82 23.0 10 F05 Female
83 22.5 12 F05 Female
84 23.5 14 F05 Female
85 20.0 8 F06 Female
86 21.0 10 F06 Female
87 21.0 12 F06 Female
88 22.5 14 F06 Female
89 21.5 8 F07 Female
90 22.5 10 F07 Female
91 23.0 12 F07 Female
92 25.0 14 F07 Female
93 23.0 8 F08 Female
94 23.0 10 F08 Female
95 23.5 12 F08 Female
96 24.0 14 F08 Female
97 20.0 8 F09 Female
98 21.0 10 F09 Female
99 22.0 12 F09 Female
100 21.5 14 F09 Female
101 16.5 8 F10 Female
102 19.0 10 F10 Female
103 19.0 12 F10 Female
104 19.5 14 F10 Female
105 24.5 8 F11 Female
106 25.0 10 F11 Female
107 28.0 12 F11 Female
108 28.0 14 F11 Female
*Mixed modeling and fixed effect test*
SAS
proc mixed data=formixed;
class gender age person;
model y = gender|age;
repeated / type=cs sub=person;
run;
output of interest to me
Tests of Fixed Effects
Source NDF DDF Type III F Pr > F
GENDER 1 25 9.29 0.0054
AGE 3 75 35.35 0.0001
GENDER*AGE 3 75 2.36 0.0781
R (nlme package)
y=lme(distance~Sex*age, random=(~1|Subject), data=dental)
anova(y)
numDF denDF F-value p-value
(Intercept) 1 75 4123.156 <.0001
Sex 1 25 9.292 0.0054
age 3 75 40.032 <.0001
Sex:age 3 75 2.362 0.0781
Now this isn't exact but it is extremely close, however when I try to
replicate the unstructured,
SAS
proc mixed data=formixed;
class gender age person;
model y = gender|age;
repeated / type=un sub=person;
run;
Tests of Fixed Effects
Source NDF DDF Type III F Pr > F
GENDER 1 25 9.29 0.0054
AGE 3 25 34.45 0.0001
GENDER*AGE 3 25 2.93 0.0532
R
either
y=lme(distance~Sex*age, random=(~1|Subject), corr=corSymm(,~1|Subject),
data=dental)
anova(y)
or
z=lme(distance~Sex*age, random=(~1|Subject), corr=corSymm(), data=dental)
anova(z)
gives the output
numDF denDF F-value p-value
(Intercept) 1 75 4052.028 <.0001
Sex 1 25 8.462 0.0075
age 3 75 39.022 <.0001
Sex:age 3 75 2.868 0.0421
What am I doing wrong to replicate the unstructured linear mixed model from
SAS?
Regards,
Charles
[[alternative HTML version deleted]]
_______________________________________________
R-sig-mixed-models at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
-----------------------------------------------------------------------
Confidentiality Notice: This e-mail message, including any attachments,
is for the sole use of the intended recipient(s) and may contain
privileged and confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply e-mail and destroy
all copies of the original message.
More information about the R-sig-mixed-models
mailing list