[BioC] using mat.samp in siggenes

[Holger Schwender]

In the unpaired case, a balanced permutation is a permutation of the group
labels such that each group contains the same number of samples from each of
the original groups. So if each of the 2 groups contain n samples, then in a
balanced permutation, each group contains n/2 samples from group 1 and n/2
samples from group 2. If it is not possible to get balanced permutations,
they should be at least as balanced as possible -- what ever this means. The
argument bal for balanced permutations which can actually only be used in
the unpaired case will be removed in the next version of siggenes.

If I understand you correctly, sam uses what you call balanced permutations
in the paired case.

Best, 
Holger

> Thanks very much for this extended reply.
> Out of curiosity, does sam() do unbalanced
> permutations when the data are paired?  If so,
> I suggest that it should default to only
> doing balanced permutations.  By 'balanced'
> in the paired case, I mean that within each pair
> the assignment to treatment/control is randomized.
> Actually, I'm not sure what 'balanced' means in the context
> of unpaired data.
> 
> Thanks again,
> Katie
> 
> ----- Original Message ----- 
> From: "Holger Schwender" <holger.schw at gmx.de>
> To: "James W. MacDonald" <jmacdon at med.umich.edu>
> Cc: <katiek at u.washington.edu>; <bioconductor at stat.math.ethz.ch>
> Sent: Thursday, September 16, 2004 2:28 AM
> Subject: Re: [BioC] using mat.samp in siggenes
> 
> 
> > In the next version of siggenes, version 1.2.x, which will (hopefully)
> be
> > part of Release 1.5 of Bioconductor, it won't be necessary anymore (for
> a
> > SAM analysis) to specify mat.samp if you wanna do complete permutation
> since
> > this will be possible by setting B=0 or to an integer equal to or larger
> > than the number of all possible permutations. However, it is still
> possible
> > to specify mat.samp if you, e.g., would like to use only balanced
> > permutations.
> >
> > Besides from this, there will be many more changes in siggenes since I
> have
> > rewritten the code for a SAM analysis. E.g., the function sam.plot is
> > replaced by the functions plot and summary, 'data' in sam(...) can be
> also
> > an exprSet object (using an exprSet object, pData can be used for
> specifying
> > cl), it is possible to do a two class unpaired analysis assuming
> *unequal*
> > variances, i.e. an analysis using Welch's t-statistic, a multiclass
> analysis
> > or an analysis of categorical data such as SNP data, respectively, it is
> > possible to add the locus links to the output, the significant genes are
> now
> > ordered by their absolute value, i.e. by their "significance", ...
> >
> > The usage of sam will, however, be almost the same -- at least for the
> > "important" arguments. I have only changed the default of med to
> med=FALSE
> 
> > which means that the mean (and not the median) number of falsely called
> > genes is computed by default.
> >
> > There also will be a manual (hopefully a vignette) which will describe
> all
> > these changes. Thanks to Kathleen, I will also add a section of how
> mat.samp
> > has to be specified. I have almost forgotten this.
> >
> > There however will be *no* changes in the empirical Bayes functions. New
> > versions of find.a0, ebam and ebam.wilc with the same features as sam
> might
> > possibly be part of Release 1.6 of Bioconductor.
> >
> > Sorry for this pretty long mail. I actually only wanted to say that
> complete
> > permutation is possible in the next version of siggenes.
> >
> > Best,
> > Holger
> >
> >
> >
> > > Kathleen Kerr wrote:
> > >
> > > > I am using the sam() function in siggenes and I
> > > > cannot find information about using the option mat.samp:
> > > >
> > > >  mat.samp: a permutation matrix. If specified, this matrix will be
> used,
> > > >         even if 'rand' and 'B' are specified.
> > > >
> > > > I have a dataset with four treated samples and four control samples
> > > > that are paired, so my class label is:
> > > > cl=c(-1,-2,-3,-4, 1,2,3,4)
> > > >
> > > > There are only 2^4=16 permutations, so I would just like to
> enumerate
> > > > them.  The helpfile doesn't give any information about what the
> > > > permutation matrix should look like.
> > >
> > > The permutation matrix should be a 16 x 4 matrix of 1's and -1's
> > >
> > > 1 1 1 1
> > > -1 1 1 1
> > > 1 -1 1 1
> > > 1 1 -1 1
> > > etc
> > >
> > > HTH,
> > >
> > > Jim
> > >
> > >
> > >
> > > >
> > > > I have tried the matrix with sixteen rows, each of which
> > > > looks something like
> > > > c(-1, 2, 3,-4, 1,-2,-3,4)
> > > > but I get an error message that this is invalid.
> > > >
> > > > _______________________________________________
> > > > Bioconductor mailing list
> > > > Bioconductor at stat.math.ethz.ch
> > > > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > >
> > >
> > > -- 
> > > James W. MacDonald
> > > Affymetrix and cDNA Microarray Core
> > > University of Michigan Cancer Center
> > > 1500 E. Medical Center Drive
> > > 7410 CCGC
> > > Ann Arbor MI 48109
> > >
> > > _______________________________________________
> > > Bioconductor mailing list
> > > Bioconductor at stat.math.ethz.ch
> > > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > >
> >
> > -- 
> > Supergünstige DSL-Tarife + WLAN-Router für 0,- EUR*
> > Jetzt zu GMX wechseln und sparen http://www.gmx.net/de/go/dsl
> >
> >
> 

-- 
Supergünstige DSL-Tarife + WLAN-Router für 0,- EUR*
Jetzt zu GMX wechseln und sparen http://www.gmx.net/de/go/dsl