Changes in version 1.4.0
* Added isoformProportions(), which can be run after
  scaleInfReps() and optionally after filtering out
  transcripts using labelKeep(). Running swish() after
  isoformProportions() will produce differential transcript
  usage (DTU) results, instead of differential transcript
  expression (DTE) results. Example in vignette.
+ Default number of permutations increased from 30 to 100.
  It was observed that there was too much fluctuation in the
  DE called set for nperms=30 across different seeds, and
  setting to 100 helped to stabilize results across seeds,
  without increasing running time too much. For further reduced
  dependence on the seed, even higher values of nperms
  (e.g. 200, 300) can be used.

Changes in version 1.3.8
* Added isoformProportions(), which can be run after
  scaleInfReps() and optionally after filtering out
  transcripts using labelKeep(). Running swish() after
  isoformProportions() will produce differential transcript
  usage (DTU) results, instead of differential transcript
  expression (DTE) results. Example in vignette.

Changes in version 1.3.4
+ Default number of permutations increased from 30 to 100.
  It was observed that there was too much fluctuation in the
  DE called set for nperms=30 across different seeds, and
  setting to 100 helped to stabilize results across seeds,
  without increasing running time too much. For further reduced
  dependence on the seed, even higher values of nperms
  (e.g. 200, 300) can be used.

Changes in version 1.2.0
+ Switching to a faster version of Swish which only
  computes the ranks of the data once, and then re-uses
  this for the permutation distribution. This bypasses
  the addition of uniform noise per permutation and
  is 10x faster. Two designs which still require
  re-computation of ranks per permutation are the
  paired analysis and the general interaction analysis.
  Two-group, stratified two-group, and the paired
  interaction analysis now default to the new fast
  method, but the original, slower method can be used
  by setting fast=0 in the call to swish().
+ Adding Rcpp-based function readEDS() written by
  Avi Srivastava which imports the sparse counts stored
  in Alevin's Efficient Data Storage (EDS) format.
+ Changed the vignette so that it (will) use a linkedTxome,
  as sometime the build would break if the Bioc build
  machine couldn't access ftp.ebi.ac.uk.
+ Add 'computeInfRV' function. InfRV is not used in the
  Swish methods, only for visualization purposes in the
  Swish paper.
+ removed 'samr' from Imports, as it required source
  installation, moved to Suggests, for optional qvalue
  calculation

Changes in version 0.99.30
+ added two interaction tests, described in ?swish
+ incorporate qvalue package for pvalue, locfdr and qvalue
+ added plotMASwish() to facilitate plotting
+ wilcoxP is removed, and the mean is used instead

Changes in version 0.99.0
+ fishpond getting ready for submission to Bioc