[R-meta] Coding for length of time in longitudinal studies given its use in meta-regression
Dear Michael and list members,
Thank you for your response. May I clarify the following:
1) We have a "continuous" moderator called `length` showing the "amount of
time" allowed between the last treatment and each post-test in each of our
longitudinal studies.
2) QUESTION: Since `length` is a moderator that only applies to
"post-tests" (see definition above) and NOT to the "pre-tests", would it be
accurate (i.e., from the perspective of using `length` in meta-regression)
to put "0" for all "pre-test" rows in each study when coding for `length`?
3) An example coded study (per your request, I'm also sharing a link for
easier access to this one example coded study). In this one example coded
study, descriptives are denoted by: mT,sdT,nTmC,sdC,nC. Also, 'group'
denotes the treatment group index, and 'time' is a time point indicator.
example_of_a_coded_study <- read.csv("
https://raw.githubusercontent.com/hkil/m/master/1.csv")
study mT sdT nT mC sdC nC group length time
13 xxx 67.590 19.260 17 51.900 28.320 20 1 0 0
14 xxx 67.590 19.260 17 51.900 28.320 20 1 0 0
15 xxx 53.110 21.730 18 51.900 28.320 20 2 0 0
16 xxx 53.110 21.730 18 51.900 28.320 20 2 0 0
17 xxx 59.350 18.390 20 51.900 28.320 20 3 0 0
18 xxx 59.350 18.390 20 51.900 28.320 20 3 0 0
19 xxx 82.530 14.330 17 52.750 23.260 20 1 0 1
20 xxx 77.500 16.070 18 52.750 23.260 20 2 0 1
21 xxx 81.250 13.000 20 52.750 23.260 20 3 0 1
22 xxx 82.650 15.470 17 63.900 18.890 20 1 7 2
23 xxx 76.780 20.670 18 63.900 18.890 20 2 7 2
24 xxx 80.050 12.090 20 63.900 18.890 20 3 7 2
On Tue, May 18, 2021 at 7:53 AM Michael Dewey <lists at dewey.myzen.co.uk>
wrote:
Dear Fred I am not sure I understand what you are doing but the usual route if you have a variable which is only meaningful in one condition is to also include a two-level factor for condition. Incidentally if people need to use your data it is much better to include it using dput() so they can read it in directly. Michael On 17/05/2021 22:50, Farzad Keyhan wrote:
Dear All, My collaborators and I want to code for the "length of time" (`length`) elapsed between the last treatment and each post-test in each of our longitudinal studies (an example is below). My question is that since `length` is a moderator that only applies to "post-tests" and NOT to the "pre-tests", is it correct (i.e., from the perspective of using `length` in meta-regression) to put "0" for all "pre-test" rows (up to row # 18)? Note that "0" is a meaningful value for `length`. For example, in the
study
coded below, "post-test 1" (i.e., time == 1) has an actual `length` of
"0".
Thank you for your assistance,
Fred
study mT sdT nT mC sdC nC
group length time
13 xxxxxxxxxxx 67.590 19.260 17 51.900 28.320 20 1 0
0
14 xxxxxxxxxxx 67.590 19.260 17 51.900 28.320 20 1 0
0
15 xxxxxxxxxxx 53.110 21.730 18 51.900 28.320 20 2 0
0
16 xxxxxxxxxxx 53.110 21.730 18 51.900 28.320 20 2 0
0
17 xxxxxxxxxxx 59.350 18.390 20 51.900 28.320 20 3 0
0
18 xxxxxxxxxxx 59.350 18.390 20 51.900 28.320 20 3 0
0
19 xxxxxxxxxxx 82.530 14.330 17 52.750 23.260 20 1 0
1
20 xxxxxxxxxxx 77.500 16.070 18 52.750 23.260 20 2 0
1
21 xxxxxxxxxxx 81.250 13.000 20 52.750 23.260 20 3 0
1
22 xxxxxxxxxxx 82.650 15.470 17 63.900 18.890 20 1 7
2
23 xxxxxxxxxxx 76.780 20.670 18 63.900 18.890 20 2 7
2
24 xxxxxxxxxxx 80.050 12.090 20 63.900 18.890 20 3 7
2
[[alternative HTML version deleted]]
_______________________________________________ R-sig-meta-analysis mailing list R-sig-meta-analysis at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
-- Michael http://www.dewey.myzen.co.uk/home.html