Random effects metagenomeSeq
2
0
Entering edit mode
PJ • 0
@pj-8012
Last seen 4.8 years ago
United States

Does metagenomeSeq handle mixed models including random effects? If yes, how can I incorporate those random effects in the model?Thanks!

model.matrix • 1.0k views
0
Entering edit mode
@joseph-nathaniel-paulson-6442
Last seen 4.6 years ago
United States

Short answer: Currently, no.

The fitZig function outputs a MArrayM object which can then be used to run limma's duplicatedCorrelation function.

This is implemented in metagenomeSeq using the duplicateCorrelation function when the option useMixedModel==TRUE, but is still considered in development.

See the instructions in limma's manual for how the function should be called.

0
Entering edit mode
noelle.noyes ▴ 30
@noellenoyes-7241
Last seen 5.7 years ago
United States

Hello,

I also have a question about use of random effects with fitZig.  I am currently trying to run the following model:

design = model.matrix(~0+var1+var2+var3)
dup <- duplicateCorrelation(MRcounts(MRexp),block=subject, design= design)
model = fitZig(obj = MRexp, mod = design, control = settings, useCSSoffset=FALSE, useMixedModel=dup$consensus) This is what I understand to be correct from limma's user manual. However, when I then call res$dupcor, I get "NULL" -- as if that part of the model wasn't called.

In addition, the results of the above model are exactly the same when I don't call "useMixedModel".

My conclusion is that fitZig isn't incorporating the "useMixedModel" statement.  Is this correct, or is something wrong with my model statement?

I am using metagenomeSeq version 1.12.0.

Thank you in advance for any help you can provide!

Noelle