|
dear S+ users,
I am using S+2000 release 1 on Windows 98.
i have a large dataset of outpatient psychotherapy patients which i
want to analyze via lme. it is unbalanced (2-4 measures per subject at
different times over a 2-year observation period). right now my major problem
is that the dose (one of three kinds of psychotherapies of different durations)
does not set in exactly at t1. is it possible to model that by simply including
as a covariate the difference in time between first observation and setting
in of dose? couldn't find anything similar in pinheiro and bates' book.
when adding other covariates (of which i have plenty), what would be the
best way to deal with that problem? below some information on the dataset.
any suggestions would be highly appreciated. thank you so much.
bernd puschner
### grouping procedure
psyther.gr <- groupedData(IMPAIR ~ DAYS | CODE, data = ""
outer = ~ T1.DOSE)
### part of dataset
### variables: code = subject, impair = global psychological impairment,
days = days from t1 (0 = t1),
### t1.dose = start of dose, ther = type of dose
> psyther.gr
Grouped Data: SCSW.010 ~ DAYS | CODE
CODE IMPAIR DAYS T1.DOSE THER
1 989001 0.57
0.0 64
2
2 989001 0.96 233.0
64 2
3 989001 0.76 362.0
64 2
4 989001 NA
557.0 64
2
5 989002 1.87
0.0 32
3
6 989002 1.14
24.0 32
3
7 989002 1.44 173.0
32 3
8 989002 0.54 550.0
32 3
9 98A001 0.40
0.0 44
3
10 98A001 0.39 174.0
44 3
11 98A001 0.13 521.0
44 3
12 98A002 1.20
0.0 420 3
13 98A002 0.99 70.0
420 3
14 98A003 0.80
0.0 390 3
15 98A003 0.40 349.0
390 3
16 98A003 0.36 567.0
390 3
17 98A004 1.22
0.0 -37 1
18 98A004 0.97 327.0
-37 1
19 98A004 0.11 565.0
-37 1
20 98A004 0.19 572.0
-37 1
21 98A005 1.43
0.0 399 3
22 98A005 0.94 184.0
399 3
23 98A005 0.53 496.0
399 3
24 98A005 1.16 548.0
399 3
25 98A007 1.16
0.0 389 1
26 98A007 0.72 124.0
389 1
27 98A007 0.33 363.0
389 1
28 98A007 0.36 662.0
389 1
...
puschner.vcf
Description: Visitenkarte für Bernd Puschner
|