Code syntax tweaks to remove warnings raised in later version of R.
Increase the forceRand unit test from 10,000 to 30,000 iterations. This solves an occasional unit test failure due to sampling.
Previously, if no random seed is specified to the harman
function, a seed of 0 was used. Now a random seed between 0 and 1e9 is
picked at random.
Dynamically resizes legends depending upon the number of batches in
pcaPlot
A custom prcomp function to get the appropriate scores. The standard R
prcomp function did not work in instances where the number of samples
(matrix columns) was greater than the number of assays (matrix rows), so a
special case is needed for less assays than samples. We need to use u'
instead of v from the SVD. Presently, this is under developement, so an
error is thrown if rows < cols.
prcompPlot now has an argument for scaling, which defaults to
FALSE. Previously, scaling was always TRUE. This new default makes
prcompPlots plots agree with the 'original' plots of plotting
harmanresults objects.
Extensive updating of the vignette with a new comparison to ComBat from sva.
First public version on Bioconductor.