[BioC] 2x3 ANOVA

[Jeff Lande]

Thanks Denise,

My initial setup was reading Gday as a covariate rather than a factor.
I changed the values of 4, 14 and 25 to "4d", "14d" and "25d" and with
the resulting coefficients, it makes a lot more sense.

Jeff

> -----Original Message-----
> From: Denise Scholtens [mailto:dscholtens at northwestern.edu]
> Sent: Thursday, July 15, 2004 1:18 PM
> To: bioconductor at stat.math.ethz.ch; land0038 at umn.edu
> Subject: Fwd: Re: [BioC] 2x3 ANOVA
> 
> 
> >Date: Thu, 15 Jul 2004 12:51:43 -0500
> >To: "Jeff Lande" <land0038 at umn.edu>
> >From: Denise Scholtens <dscholtens at northwestern.edu>
> >Subject: Re: [BioC] 2x3 ANOVA
> >
> >>
> >>
> >>Now, I'd like to start running some contrasts to see the effects at
> >>particular graft type/day combinations (ie 14 day allografts) and
also
> >>look at fold changes.  I think both of these functions are possible
> >>within the factDesign package, but I'm not sure how to input the
second
> >>and the third parameters in the par2lambda function.  If there's
anyone
> >>who can give me some direction on this, I would greatly appreciate
it.
> >
> >Hi Jeff,
> >
> >par2lambda takes three coefficients - "betaNames", "betas", and
"coefs".
> >
> >"betaNames" refers to the names of the coefficients that are returned
> with
> >the lm, for example, you could do:
> >
> >betaNames <- names(lm.full[["coef"]])
> >
> >"betas" and "coefs" are used to define the specific contrasts of
> >interest.  The idea of the par2lambda function is not to have to
> construct
> >a contrast matrix by hand  - if you did have to do this, you would
need
> to
> >look at the order of the coefficients returned by the lm and make
sure
> >your contrast matrix is specified accordingly.  When performing many
> >contrasts on several different factors, this can be a pain.  In any
> event,
> >suppose your lm returns coefficients named "beta1", "beta2", "beta3",
and
> >"beta4" and you want to test the contrast beta2 - beta4=0.  Then you
> would
> >specify
> >
> >betas <- list(c("beta2","beta4"))
> >coefs <- list(c(1,-1))
> >
> >and you're all set to use the par2lambda function.  The list
structure is
> >helpful if you want to simultaneously test two or more contrasts.
> Suppose
> >in addition to beta2-beta4=0, you want to test beta3+beta4=0.  Then
you
> >would say
> >
> >betas <- list(c("beta2","beta4"),c("beta3","beta4"))
> >coefs <- list(c(1,-1),c(1,1))
> >
> >par2lambda(betaNames,betas,coefs) will then return the appropriate
> >contrast matrix for use with contrastTest or findFC.  There are
examples
> >of how to specify these parameters in the documentation example for
the
> >par2lambda function and also in the factDesign vignette.
> >
> >Hope this helps -
> >Denise
> >
> >
> >
> >>
> >>Thanks,
> >>
> >>Jeff Lande
> >>Graduate Student - University of Minnesota
> >>
> >>
> >>         [[alternative HTML version deleted]]
> >>
> >>_______________________________________________
> >>Bioconductor mailing list
> >>Bioconductor at stat.math.ethz.ch
> >>https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
> >
> >******************************************
> >Denise Scholtens, Ph.D.
> >Research Assistant Professor
> >Northwestern Medical School
> >Department of Preventive Medicine
> >680 N. Lake Shore Drive Suite 1102
> >Chicago, IL 60614
> >phone: 312.503.7261
> >fax: 312.908.9588
> >dscholtens at northwestern.edu
> 
> ******************************************
> Denise Scholtens, Ph.D.
> Research Assistant Professor
> Northwestern Medical School
> Department of Preventive Medicine
> 680 N. Lake Shore Drive Suite 1102
> Chicago, IL 60614
> phone: 312.503.7261
> fax: 312.908.9588
> dscholtens at northwestern.edu