[R-meta] Redundant predictors

[Viechtbauer, Wolfgang (SP)]

Dear Arne,

I would have to see your data/model, but yes, if the model matrix is not of full rank (which could be because certain combinations of two factors do not occur in a model that allows the two factors to interact), then rma.mv() will drop predictors so the model matrix is of full rank. Here is an example with two factors:

set.seed(1234)
yi <- rnorm(9)
F. <- factor(c("a","a","a","b","b","b","c","c","c"))
G. <- factor(c("1","2","2","1","2","2","1","1","1"))
vi <- rep(1, 9)
rma.mv(yi, vi, mods = ~ F.*G.)

Note that the combination c-2 never occurs. So, the interaction term for this term is dropped from the model.

Is this a problem? Well, it does mean that we cannot estimate certain coefficients. In the example above, we cannot test whether the difference between F.c and F.a is different for the two levels of factor G (essentially, the difference between F.c and F.a is assumed to be the same for both levels of factor G, that is, 0.6451). But this aside, the model is still ok.

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On Behalf Of Arne Janssen
Sent: Monday, 04 February, 2019 16:16
To: 'r-sig-meta-analysis using r-project.org'
Subject: [R-meta] Redundant predictors

Dear Wolfgang,

I am running an rma.rv with two categorical and one continuous moderator 
plus their interactions. Analysis of the full model give a Warning 
message: In rma.mv ( ...)  : Redundant predictors dropped from the model.
Based on the model output, I guess this is caused by the absence of all 
combinations of the two categorical moderators, is that correct? If so, 
is this something to worry about?

Best wishes,
Arne