[R-sig-ME] Warning in rankMatrix(X)
Ben Bolker
bbolker at gmail.com
Thu Oct 17 16:40:28 CEST 2013
Juan Kamienkowski <jkamienk at ...> writes:
>
> Hi,
>
> I moved to lmer 1.0-4 and R 3.0.2. Now lmer() and glmer() functions are
> working much slower than before, and I get the following warning message,
> which I suspect is related to that.
>
> Warning in rankMatrix(X) :
> rankMatrix(<large sparse Matrix>, method = 'tolNorm2') coerces to dense
> matrix.
> Probably should rather use method = 'qrLINPACK' !?
>
> Do you have any advice on this? Where can I change the method of the
> rankMatrix()?
>
This is a known bug, actually in Matrix: we hope to fix it soon,
but it will require a coordinated release of Matrix, RcppEigen,
and lme4. You can ignore it for now, or use suppressWarnings() if
it's really annoying you (although that will suppress *all* warnings,
so it may not be a good idea).
Sorry about things slowing down: can you tell us more about your
use case? Have you tried switching from the default Nelder_Mead to
the (sometimes faster) bobyqa setting? (see ?lmerControl)
Ben Bolker
More information about the R-sig-mixed-models
mailing list