I have a performed a repeated measures ANOVA. My problem is that if I
want to get the adjusted means for the three-way interaction (it isn't
significant in this example but assume it was) diab*currshoulder*time
using the function adjust it produces a blank table as shown below.
However I understand it has something to do with my between-subjects
error term labelled uid. How can I get my 3 way interaction means at the
average age but also at the 'average' uid/subject(centering the uid
effect at zero)?
I have attached the example dataset.
Thanks,
Janine.
. anova spadi age diab currshoulder diab*currshoulder /
uid|diab*currshoulder time diab*time currshoulder*ti
> me diab*currshoulder*time, continuous(age) repeated(time) sequential
Number of obs = 558 R-squared =
0.8701
Root MSE = 11.2057 Adj R-squared =
0.7620
Source | Seq. SS df MS F
Prob > F
----------------------+----------------------------------------------------
Model | 255678.229 253 1010.58589 8.05
0.0000
|
age | 4830.84766 1 4830.84766 7.25
0.0076
diab | 8418.50341 1 8418.50341 12.63
0.0005
currshoul~r | 75854.8577 1 75854.8577 113.83
0.0000
diab*currshoul~r | 1118.32523 1 1118.32523 1.68
0.1964
uid|diab*currshoul~r | 160593.71 241 666.363943
----------------------+----------------------------------------------------
time | 3988.74315 2 1994.37158 15.88
0.0000
diab*time | 402.488385 2 201.244192 1.60
0.2031
currshoul~r*time | 295.765196 2 147.882598 1.18
0.3094
diab*currshoul~r*time | 174.987866 2 87.493933 0.70
0.4990
|
Residual | 38172.7318 304 125.568197
----------------------+----------------------------------------------------
Total | 293850.961 557 527.560073
Between-subjects error term: uid|diab*currshoul~r
Levels: 246 (241 df)
Lowest b.s.e. variable: uid
Covariance pooled over: diab*currshoul~r (for repeated variable)
Repeated variable: time
Huynh-Feldt epsilon =
0.9959
Greenhouse-Geisser epsilon =
0.9719
Box's conservative epsilon =
0.5000
------------ Prob > F
------------
Source | df F Regular H-F G-G
Box
----------------------+----------------------------------------------------
time | 2 15.88 0.0000 0.0000 0.0000
0.0001
diab*time | 2 1.60 0.2031 0.2032 0.2038
0.2075
currshoul~r*time | 2 1.18 0.3094 0.3093 0.3085
0.2795
diab*currshoul~r*time | 2 0.70 0.4990 0.4984 0.4951
0.4052
Residual | 304
----------------------+----------------------------------------------------
. adjust age, by(diab currshoulder time)
------------------------------------------------------------------------------------------------------------
Dependent variable: spadi Command: anova
Variable left as is: uid
Covariate set to mean: age = 59.180023
------------------------------------------------------------------------------------------------------------
--------------------------------------------------
| time! and currshoulder
| ---- 0 --- ---- 6 --- --- 12 ---
diab | 0 1 0 1 0 1
----------+---------------------------------------
0 |
1 |
--------------------------------------------------
Key: Linear Prediction
.
Attachment:
example.dta
Description: application/unknown-content-type-stata9data
