[R-sig-ME] [Lme4-authors] lme4 question

[Martin Maechler]

Dear Alfonso,

>>>>> Alfonso  <Alfonso.Buil at unige.ch>
>>>>>     on Tue, 16 Aug 2011 16:03:21 +0200 writes:

 [ to the lme4-authors at r-forge... address ... ]

    > Hello, My names is Alfonso Buil and I work as a postdoc at
    > the University of Geneva (Switzerland).  

as I am from Zurich,  I feel some obligation to answer you ...

    > I work in
    > genomics and I have large datasets. Now, I am working in a
    > project with family data and I want to use a Linear Mixed
    > Model to account for the familiar relationships among
    > individuals. My problem is that I have to perform several
    > millions of test and that takes a while.  

"test"s in the sense of hypothesis test ?

    > My questions
    > are:
    >  - How efficient is the implementation of lme4? 
well, in comparison with what?
Some of its large scale computations (large sparse Z) are very
efficient ...

    >  - Are the computationally heavy functions written in C or
    >   FORTRAN?  
yes

    > - Is there any way to make lme4 run faster?
yes.. 
there is "always".

I'm CCing my reply to the R-SIG-Mixed-Models mailing list, to
which you really should subscribe (*and* have posted your
questions ..).

But before anyone can help you further you definitely need to
give much more details,
about the data (sizes.. ) and the models you are fitting
(exact formula, output, ...)


    > Congratulations for this amazing package.  Thanks a lot
    > for your help,

you are welcome.
Martin Maechler, ETH Zurich

    >    alfonso _______________________________________________