[R-meta] Any limitations in random-effect specification in rma.mv()?
Timothy MacKenzie
|@w|@wt @end|ng |rom gm@||@com
Sat Jun 26 01:54:20 CEST 2021
I have a set of multivariate ('outcome'), longitudinal studies ('time')
that occasionally have used multiple independent samples ('sample') of
subjects and/or more than one control group ('control'). My data
structure looks like:
study outcome time sample control obs
1 1 1 1 1 1
1 1 2 2 1 2
1 2 1 1 2 3
1 2 2 2 2 4
2 1 1 1 1 5
2 1 2 1 2 6
3 1 1 1 1 7
3 2 1 2 1 8
I intend to empirically test all possible sources of random variation in
the true effects due to 'study', 'outcome', 'sample', 'control', and
perhaps observations (i.e., each row) differences.
I know I can specify random effects for 'outcome' within studies and 'time'
within study within 'outcome' as follows:
`random = list(~ outcome | study, ~ time | interaction(study,outcome))`
But I was wondering how I could add the random effects for 'sample',
'control', and 'obs'? Ex. Is the following legitimate? Do any of 'control'
or 'sample' need to be crossed effects?
`random = list(~ outcome | study, ~ time | interaction(study,outcome), ~ 1
| sample, ~ 1 | control, ~ 1 | obs)`
Thanks, Tim
[[alternative HTML version deleted]]
More information about the R-sig-meta-analysis
mailing list