[R-sig-ME] incorporate pedigree into lmer
Douglas Bates
bates at stat.wisc.edu
Tue Feb 6 19:04:37 CET 2007
Sorry that I haven't been part of this conversation until now. We had
a weather-related telephone failure in our city that knocked out voice
and Internet communications to our home for a few days. I have only
been able to respond to email from the office and what with meetings,
classes, etc. I have fallen behind.
The questions that Jon is asking relate to the particular form of
random effects models that animal scientists use taking into account
the pedigree of the animals. These are an example of a model with
"carry-over" in the random effects. The random effect for an animal
enters into the model for that animal's responses and for all of that
animal's progeny's responses.
The current form of lmer does not allow for such carry-over. The
models that can be fit are restricted to what might be called
contemporaneous random effects. In particular, only one level of a
grouping factor can be involved in the model for a given response.
That's the bad news. The good news is that there is nothing in the
computational methods that precludes carry-over in the model. It is
"simply" a matter of modifying the model matrix for the random effects
before creating the mer objects. The pedigree class was defined
exactly for this purpose. If you look at
example("pedigree-class")
you will see that there are methods for creating the matrices to be
used for the transformation but that work is not complete.
On 2/6/07, Doran, Harold <HDoran at air.org> wrote:
> It's good to keep this on list for others to chime in and for reference. You need the ranef extractor
>
> ranef(fm1)$sire
>
> > -----Original Message-----
> > From: Jon Hallander [mailto:Jon.Hallander at genfys.slu.se]
> > Sent: Tuesday, February 06, 2007 10:41 AM
> > To: Doran, Harold
> > Subject: RE: [R-sig-ME] incorporate pedigree into lmer
> >
> > Ok, thanks! This seems to work, but does it utilize the
> > pedigree information? When I type the command summary(fm1) I
> > do not get individual random effect values (breeding values),
> > but variances of the random effects:
> >
> > > fm1 <- lmer(y ~ 1 +(1|sire) + (1|dam), data)
> > > summary(fm1)
> > Linear mixed-effects model fit by REML
> > Formula: y ~ 1 + (1 | sire) + (1 | dam)
> > Data: data1
> > AIC BIC logLik MLdeviance REMLdeviance
> > 28.24 27.62 -11.12 23.45 22.24
> > Random effects:
> > Groups Name Variance Std.Dev.
> > sire (Intercept) 1.75e-09 4.1833e-05
> > dam (Intercept) 1.75e-09 4.1833e-05
> > Residual 3.50e+00 1.8708e+00
> > number of obs: 6, groups: sire, 4; dam, 3
> >
> > Fixed effects:
> > Estimate Std. Error t value
> > (Intercept) 3.5000 0.7638 4.583
> >
> >
> >
> > /Jon
> >
> > -----Original Message-----
> > From: Doran, Harold [mailto:HDoran at air.org]
> > Sent: den 6 februari 2007 15:03
> > To: Jon Hallander; r-sig-mixed-models at r-project.org
> > Subject: RE: [R-sig-ME] incorporate pedigree into lmer
> >
> > Well, I'm still guessing at what you want as a fixed effect
> > and what will be random, but here is a shot
> >
> > (fm1 <- lmer(weight ~ 1 +(1|Sire) + (1|Dam), data))
> >
> >
> >
> > > -----Original Message-----
> > > From: Jon Hallander [mailto:Jon.Hallander at genfys.slu.se]
> > > Sent: Tuesday, February 06, 2007 8:26 AM
> > > To: Doran, Harold; r-sig-mixed-models at r-project.org
> > > Subject: RE: [R-sig-ME] incorporate pedigree into lmer
> > >
> > > Oh, sorry for being unclear. I am using a mixed linear
> > model including
> > > both random and fixed effects:
> > > y = Xb + Za + e
> > > where the vector y are observed individual data, b is a vector of
> > > fixed effects, a is an individual additive random effect that are
> > > normally distributed vectors with (co)variance A*var(a). X
> > and Z are
> > > incidence matrices relating the fixed and random effects (b and a)
> > > with the observations y. Furthermore, e is the error term,
> > w normally
> > > distributed with mean zero and variance var(e). A is the additive
> > > genetic relationship matrix which is computed using the pedigree
> > > information (given below). A typical data might look like this:
> > > Id Sire Dam y (i.e. weight in kg)
> > > 1 Na Na 4.3
> > > 2 Na Na 4.7
> > > 3 1 2 4.4
> > > 4 1 Na 4.1
> > > 5 4 2 4.3
> > > 6 4 2 4.2
> > >
> > > Thanks for the help.
> > >
> > > Best regards,
> > > Jon
> > >
> > > -----Original Message-----
> > > From: r-sig-mixed-models-bounces at r-project.org
> > > [mailto:r-sig-mixed-models-bounces at r-project.org] On Behalf
> > Of Doran,
> > > Harold
> > > Sent: den 6 februari 2007 13:51
> > > To: Jon Hallander; r-sig-mixed-models at r-project.org
> > > Subject: Re: [R-sig-ME] incorporate pedigree into lmer
> > >
> > > Jon
> > >
> > > I think you will find the members of this list helpful, but
> > you have
> > > given zero information that we can use to help you with
> > your problem.
> > > Please provide a representation of your model and a description of
> > > your data. Or better yet, some sample data. I don't have
> > any idea what
> > > an "animal model" is (maybe others do), but if you can provide a
> > > statistical representation, maybe we can walk you through this.
> > >
> > > Harold
> > >
> > > > -----Original Message-----
> > > > From: r-sig-mixed-models-bounces at r-project.org
> > > > [mailto:r-sig-mixed-models-bounces at r-project.org] On
> > Behalf Of Jon
> > > > Hallander
> > > > Sent: Tuesday, February 06, 2007 7:07 AM
> > > > To: r-sig-mixed-models at r-project.org
> > > > Subject: [R-sig-ME] incorporate pedigree into lmer
> > > >
> > > > Hello,
> > > >
> > > > I am a novice using the lme4 library and would like some help
> > > > regarding the function lmer. I would like to solve the
> > animal model
> > > > using several random effects (e.g. individual (animal) and
> > > dominance
> > > > effects). Also, I would like to infer the additive and dominance
> > > > relationship matrices (covariance matrices). The function
> > > pedigree can
> > > > set up the pedigree structure, but how do I use it in the lmer
> > > > function? I have not found any convenient example.
> > > > Thanks in advance.
> > > >
> > > > Best regards,
> > > > Jon Hallander
> > > >
> > > > ____________________________________________
> > > >
> > > > Jon Hallander
> > > > Department of Forest Genetics and Plant Physiology Swedish
> > > University
> > > > of Agricultural Sciences
> > > > SE-901 83 UMEÅ
> > > > SWEDEN
> > > > Phone: +46 90 786 82 80
> > > > Mobile: +46 70 220 08 79
> > > > Fax: +46 90 786 81 65
> > > > E-mail: jon.hallander at genfys.slu.se
> > > > ____________________________________________
> > > > ###########################################
> > > >
> > > > This message has been scanned by F-Secure Anti-Virus for
> > > > Mic...{{dropped}}
> > > >
> > > > _______________________________________________
> > > > R-sig-mixed-models at r-project.org mailing list
> > > > https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> > > >
> > >
> > > _______________________________________________
> > > R-sig-mixed-models at r-project.org mailing list
> > > https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
> > > ###########################################
> > >
> > > This message has been scanned by F-Secure Anti-Virus for Microsoft
> > > Exchange.
> > > For more information, connect to http://www.f-secure.com/
> > >
> > ###########################################
> >
> > This message has been scanned by F-Secure Anti-Virus for
> > Microsoft Exchange.
> > For more information, connect to http://www.f-secure.com/
> >
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
More information about the R-sig-mixed-models
mailing list