May 15, 2016
View0af115ed

Version bump to trigger package rebuilding now that purl()'ing issue has been correctly identified. knitr does not create purl()'ed (Stangle equivalent) .R files if _R_CHECK_TIMINGS_ is set, which the build system was setting. Now it's not set, so these .R files are now created. See https://github.com/yihui/knitr/issues/1212 for more.

Dan Tenenbaum authored on 15/05/2016 20:18:19
May 11, 2016
View636f9f31

Bump version of all packages that use knitr for vignettes. This is because of an issue (now fixed) in knitr which failed to create purl()'ed R files from vignette sources and include them in the package. This version bump will cause these packages to propagate with those R files included.

Dan Tenenbaum authored on 11/05/2016 23:15:25
May 3, 2016
View7aadd80a

Creating branch for BioC 3.3 release.

Dan Tenenbaum authored on 03/05/2016 21:26:31
View0be58c12

Bump versions prior to creation of 3.3 branch

Dan Tenenbaum authored on 03/05/2016 21:20:18
March 7, 2016
Viewb4358efe

- Density plot made by plotFeatureClasses also now uses coord_cartesian to restrict the axes of the plot, not the data itself.

Dario Strbenac authored on 07/03/2016 23:00:05
Viewfb3a8c0e

- Performance, ranking, and selection plots used ggplot2's ylim function, which discards data and then summarises and plots it. It now uses coord_cartesian, so that all data values are kept for the plot.

Dario Strbenac authored on 07/03/2016 06:00:08
March 2, 2016
Viewb19b4d0b

- plotFeatureClasses now uses grid.newpage to avoid overlaid plots when the density and stripchart are both plotted. - plotFeatureClasses aligns density and stripchart axes. - Default margins same for all plotting functions.

Dario Strbenac authored on 02/03/2016 01:00:09
December 21, 2015
Viewfe45bd77

- previousSelection now correctly handles unmatched features in the validation dataset and extracts the necessary feature names from the ClassifyResult object. - runTest and runTests now gracefully handle errors that come from transformation, selection, traning, or prediction functions, so that the cross-validation does not stop if there is an error in any one cross-validation.

Dario Strbenac authored on 21/12/2015 04:00:06
December 16, 2015
View69bc80e4

- Calculation of log likelihood ratio corrected.

Dario Strbenac authored on 16/12/2015 06:00:11
December 3, 2015
View4896b1af

- Missing package require in SelectParams example added.

Dario Strbenac authored on 03/12/2015 04:00:10
December 2, 2015
Viewb68eba54

- plot argument to plotFeatureClasses is now used.

Dario Strbenac authored on 02/12/2015 07:00:09
December 1, 2015
View379a42b6

- subtractFromLocation gains crucial absolute argument. - Format of logical values in help pages made consistent.

Dario Strbenac authored on 01/12/2015 07:00:11
November 27, 2015
Viewd9af9ec8

- Selection frequency calculation of distribution function is corrected. - distribution gets more options to customise the plot and returned values.

Dario Strbenac authored on 27/11/2015 06:00:14
November 26, 2015
Viewb1d72aed

- performancePlot is now compatible with ggplot2 1.0.1, which doesn't handle NULLs as well as the upcoming 1.1.0 version, but Hadley is taking a long time to release it.

Dario Strbenac authored on 26/11/2015 05:00:16
November 20, 2015
View93249a5f

- Default bandwidth algorithm changed to nrd0 because SJ doesn't work with repeated values and nrd0 is faster to compute. - performancePlot previously ignored the selectionName setting for boxFillColouring and boxLineColouring. It now works. - ROCplot no longer ignores the showAUC variable. - ROCplot no longer ignores line colouring by selection method option. - performancePlot has option to show or hide legend, rather always than suppressing it. - facetting in rankingPlot, selectionPlot, and performancePlot now use reformulate to create the formula from strings, to be compliant with ggplot2 1.1.0. - runTest incorrectly determined whether a list of selection varieties were returned from .doSelection when ensemble feature selection was performed. It now checks the chosen features variable, which is not always NULL when ensemble seelction is done. - runTests now correctly handles classifications using ensemble feature selections.

Dario Strbenac authored on 20/11/2015 03:00:25