2011-02-18  Gary Lipton  <gl37@tack.isds.duke.edu>
	*  (SUMMARY) Tweaks for publishing on CRAN.

2011-01-12  Gary Lipton  <gl37@tack.isds.duke.edu>
	*  (SUMMARY) Incorporated C function bayesglm_fit() from BAS.
	
	*  converge.EMC.R

	*  fit.EMC.R

	*  post.prob.R 

2010-10-15  Gary Lipton  <gl37@tack.isds.duke.edu>

	*  Gene.EMC.R: Case-control status may be expressed as a factor,
	a logical, or a numeric.

2010-08-18  Gary Lipton  <gl37@tack.isds.duke.edu>

	*  (SUMMARY) Correctly handle case with no forced variables.

	*  combine.EMC.R: New file. This function combines results from 
	multiple runs.

	*  expand.data.SNP.R: Handle case with no forced variables.

	*  fit.EMC.R: Default is now no forced variables.

	*  Gene.EMC.R: Dynamically report library versions using
	installed.packages(). Tweaks to log file.

2010-07-07  Gary Lipton  <gl37@tack.isds.duke.edu>

	*  (SUMMARY) Argument and data checking have been added. Minor
	performance improvements have been incorporated.
	
	*  MISA is now compatible with BAS 0.9.
	
	*  Each run creates a time-stamped log file for errors and
	warnings.
	
	*  All input parameters for Gene.EMC are checked for type, range, 
	and consistency.
	
	*  The input data frame is processed in the following respects:
	   a. Monomorphic SNPs are deleted.
	   b. Case values and SNP genotypes must be in {0, 1} and
	      {0, 1, 2}, respectively.
	   c. Definitions of minor & major alleles are swapped if 
	      Nmajor < Nminor.

	*  An optional progress bar has been added.

	*  Function name changes:
	     emc.converge -> converge.EMC
	     data.snp     -> expand.data.snp
	     fit          -> fit.EMC	

	*  fit.EMC.R: Eliminated use of model.matrix() & model.frame()
	(for performance). Replaced glm.fit() with BAS::bayesglm.fit().

2010-05-20  Gary Lipton  <gl37@tack.isds.duke.edu>

	*  (SUMMARY) The case variable no longer must be called "case",
	but it must be the first column in the design matrix. A GLM
	procedure from the BAS package coded in C replaces stats::glm(). A
	warning is printed when the design matrix is not full-rank.

2010-03-09  Gary Lipton  <gl37@tack.isds.duke.edu>

	*  (SUMMARY): Speed improvements and support for parallel
	processing (multicore package) have been added. Bf4assoc has
        been added.
	
	*  crossover.R: Replaced two calls to fit() with
	lapply()/mclapply(). Simplified logic of swapping code and
	eliminated calls to cbind().
	
	*  Gene.EMC.R: Added multicore support: calls to mutation() for
	each chain may be run in parallel. Pop.emc.matrix (formerly
	pop.emc) is pre-allocated to avoid multiple calls to
	rbind(). Added column to pop.emc.matrix to record number of
	acceptances for each model. Replaced 'paste(samp.model, collapse="")'
	with more efficient 'rawToChar(as.raw(samp.model + 48))'.

	*  mutation.R: Eliminate redundant call to fit() for the current
	model, reducing calls to fit() by 25%.

	*  bf4assoc.R: New command to calculate Bayes factors in favor of
	three genetic models of association.

	*  Added vignette.
	

