Skip to content

[Bioc-devel] vbmp package announcement

3 messages · Nicola Lama, Richard Pearson, Martin Morgan

Original
Nicola Lama
# 
Hi everybody,

"vbmp" version 1.5.2 is on Bioconductor.
  
This package implements code for multiclass classification by means of
Variational Bayesian Multinomial Probit regression with Gaussian Process
priors.
This method was originally proposed by Girolami M and Rogers S (Neural
Computation 2006; 18, 1790-1817).
The package contains code examples, one which mimics the analysis described
in the paper by Zsofia Kote-Jarai et al. (Clin Cancer Res
2006;12(13);3896-3901). The related dataset is included in the package. 
Should you need any more information please do not hesitate to contact me.

Best,

 - Nicola 

-------------------------------------------------------- 
Dr Nicola Lama, PhD
Medical Statistics Unit
Department of Medicine and Public Health
Second University of Napoli
Via Luciano Armanni 5, 
80138 Napoli, Italy
Tel/Fax +39-081-5666021
1 day later
Richard Pearson
# 
Hi

I think a recent change to Biobase has caused my package (puma) to fail 
to build in the release branch when previously it was fine. Could 
someone confirm that this could indeed be the case? I want to be sure of 
this because I have recently had a paper rejected and the first reason 
given was that the reviewers found that they could not get the package 
to work. I think I have now fixed the problem but I would like to 
challenge the rejection decision saying it was due to something which 
was beyond my control.

I believe the change that caused this was committed with the following 
message on svn:

      Harmonize dimnames of ExpressionSet assayData elements, if
      possible
   
      * 'harmonize' means to ensure that all dimnames are
      consistent with names from phenoData, featureData
   
      * Only possible if elements differ with NULL dimnames, not
      if elements have different dimnames; in this case, signal an
      error about conflicting dimnames

I had an example which created an ExpressionSet which had different 
column names for different assayData elements, which previously caused 
no problems.

This leads me to wonder whether the second "rule" for updating the 
release branch (http://wiki.fhcrc.org/bioc/HowTo/Update_Release_Branch) 
needs to be tightened to include something like:

2. It is very important not to change the signature of any functions, 
not to add or remove functions, /and not to change functions such that 
new errors might be given by the function/.

Best regards

Richard.
Martin Morgan
# 
Hi Richard --

First, let me apologize for not announcing the change I made to the
release branch of Biobase.

The change (on July 10) was in response to a user bug report.

  https://stat.ethz.ch/pipermail/bioconductor/2007-July/018189.html

I was faced with the dilemma of allowing Biobase to continue to
produce data objects that were not as intended, or correcting the
problem at the risk of breaking other packages.

We will work to keep changes to the release branch to a minimum, and
to diligently report these changes to bioc-devel in the future.

Please let me know (off-list) if you would like something more formal
to include with your challenge.

Best,

Martin

Richard Pearson <richard.pearson at postgrad.manchester.ac.uk> writes: