Change Log for the sigPathway package

* Version 1.9.2
	- Bug Fix
	  - added code to getPathwayStat.R such that the functions
	    do not fail when processing newer annotation packages (e.g.,
	    hgu133.db) with ".db" in their name

* Version 1.5.2
	- Bug Fix
	  - changed one line in .convertDF2HTML() of writeHTML.R to
	    repair HTML layout bug that takes place when the number of rows
	    is smaller than the number of columns in the input
	    data frame

* Version 1.5.1
	- Changes
	  - clarified the use of the "annotpkg" option in
	    getPathwayStatistics.Rd and getPathwayStatistics.NGSk.Rd
	- Additions
	  - importGeneSets.R: collection of functions to import gene sets
	    stored in GMT, GMX, GRP, and XML formats
	  - importGeneSets.Rd: help file accompanying importGeneSets.R

* Version 1.5.0
	- Changes
	  - Version numbering bump for BioConductor 2.1 development

* Version 1.3.1
	- Changes
	  - added "latin1" encoding descriptor to one of the pathway titles
	    in the example data set to remove warning from R CMD check,
	    e.g., the character vector in G[[617]]$title

* Version 1.3.0
	- Changes
	  - Version numbering bump for BioConductor 2.0 development

* Version 1.2.0
	- Changes
	  - Version numbering bump for BioConductor 1.9 release

* Version 1.1.7
	- Changes
	  - added $(FLIBS) to Makevars (to remove warnings from
	    R 2.4.0 alpha 20060906 version)
	  - added check for ENTREZID in Bioconductor annotation packages
	    (see getPathwayStat.R) because LOCUSID is marked as deprecated
	    for Bioconductor 1.9 annotation packages

* Version 1.1-6
	- Changes
	  - switched the ordering of the means in the 2 sample t-statistic
	    calculations to match the convention used in multtest and other
	    statistical packages
	  - because of the above change, the elements within the "phenotype"
	    vector of the example data are changed from "NORM" and "IBM" to
	    "0_NORM" and "1_IBM", respectively
	  - updated the vignette to reflect the changes in the "phenotype"
	    vector and the sign convention of the t-test
	  - fixed inconsistencies in the vignette between the R output and
	    the pre-generated PNG figures
	  - fixed minor typographical errors in the vignette and the help files
	
* Version 1.1-5
	- Changes
	  - replaced the full inflammatory myopathy example data set with
	    a subset to reduce overall file size of package and to satisfy
	    Bioconductor package size requirements
	  - clarified authorship of R package

* Version 1.1-4
	- Bug Fix
	  - adjusted the code so that if the user flips the phenotypes,
	    the results will stay consistent with each other, even when
	    the number of unique permutation is small (e.g.,
	    phenotype = c(0,0,0,1,1,1) has 20 possible unique permutations.
	    we compare results from the original phenotype to the 19 other,
	    permuted phenotypes)
	  - adjusted the code so that when the permuted phenotype is equal
	    to the unpermuted phenotype, the phenotype is repermuted
	- Changes
	  - replaced the .First.lib() in zzz.R with NAMESPACE, 
            R's name space management system
	  - split the R and C codes into several files for easier 
	    maintainability and addition of future features
	  - modified the R code to reflect changes made to the parameter calls 
	    in the C code
	  - modified examples in package documentation to illustrate how one
	    can use a Bioconductor annotation package (e.g., hgu133a) to
	    make the results from getPathwayStatistics() more informative
	  - modified gene set functions so that the phenotypes do not have to
	    be ordered (done for the purpose of reproducibility)
	  - modified t_R() and their variants to have the same parameter calls
	  - replaced t_null_R() and their variants with null_c() and null_R()
	- Additions
	  - when ngroups > 2, the f-statistic (from one-way ANOVA, which 
	    assumes equal variances among different groups) is calculated
	  - created functions to write results of pathway analysis to HTML
	    files for the user's benefit
	  - created functions to check users' inputs and to reduce
	    code redundancy
	  - for situations where the number of random permutations is 
	    less than the number of complete permutations, the program will
	    use complete permutations (by lexicographical order)
	    unless specified otherwise

* Version 1.1-3
	- Bug Fix
	  - in rare cases, the weight calculation routine could get trapped
	    in an infinte loop.  The initial value of lambda is now set to
	    either 0.001 or the smallest eigenvalue greater than 0.001.
	  - lambda is sometimes not converging to minimize the smallest weight
	    of a gene set.  This was mitigated with more strigent if() and
	    while() statements.
	  - calculate.NEk() is not reporting the correct intermediate values
	    for t.set (although the much more important and meaningful
	    t.set.new is correctly calculated).  Intermediate values of t.set
	    (for those who need it) are adjusted in the R code.
	- Changes
	  - revised calc_GSEA() in C source code so that it would only use
	    sparse matrices.  All relevant functions in C now rely on
	    sparse matrices.
	  - replaced calc_NTk() with calc_NGSk() to let the user specify
	    his vector of test statistic for gene label permutations; NTk
	    calculations are still handled through calculate.NTk() in R
	  - made minor grammatical changes to several documentation files
	- Additions
	  - added a sentence in calcTStatFast() R documentation to clarify
	    the expected sign of the t-statistic in the 2-group comparison
	  - getPathwayStatistics() now allows the user to specify a BioConductor
	    annotation package to fetch the probes' accession numbers, Entrez
	    Gene IDs, gene names, and gene symbols.
	  - new function: calc_NGSk() in C and calculate.NGSk() in R.  Both
	    functions take in a vector of test statistic for each probe set
	    for its gene label permutations.  This function is more generalized
	    than previous versions of calculate.NTk() because the user can now
	    specify test statistics from other association measures (e.g.,
	    Wilcoxon)
	  - new function: rankPathways.NGSk() works like rankPathways() except
	    the ranking is performed only on one set of pathway analysis results
	  - new function: getPathwayStatistics.NGSk() works like
	    getPathwayStatistics() except the arguments take in a vector of
	    probe test statistics instead of a matrix of expression values and
	    a phenotype vector

* Version 1.1-2
	- Bug Fix
	  - isinf() was replaced with R_FINITE() to account for cases where
	    some compilers (e.g., in Solaris 2.3) do not define isinf()
	- Changes
	  - revised remove_zero_cols() in C source code so that one does not
	    need to expand the sparse matrix to remove the "zero" columns

* Version 1.1-1
	- Bug Fix
	  - repaired typos that resulted in using the wrong phenotypes in 
	    the 1-sample calculations
	- Changes
	  - reimplemented the ranking and q-value code for faster calculations;
	    the number of permutations specified in the example code was
	    increased from 100 to 1000 to show this change in speed and to
	    make the results from the examples more reproducible
	  - modified the rankPathways function so that (1) it now tells the
	    percent of probe sets in the pathway that have greater mean in
	    1 sample compared to another sample and (2) it also allows the
	    user to specify whether to summarize the top N pathways from each
	    statistic or the top N pathways when referring to all pathways
	  - consequently, the input arguments for the rankPathways function
	    has been revamped; the example code was updated with the newer
	    set of input arguments
	  - slight code cleanup to remove redundant lines (e.g., the
	    calcTStatFast function checked twice to see whether the ngroups
	    variable contain 1 or 2)

* Version 1.1-0
	- First public release
