Upcoming version 6.2

[Kim-Anh Le Cao]

• fix bitbucket with KA

• create a package dependency mixOmics-data

• seems that perf.plsda does not run, possibly in the case where nzv = TRUE

• plotLoadings for PCA

• plot.tune legend as a side legend ( mar(..))

• double check plotIndiv(style = '3d') with a col input (some issues during mixOmics workshop)

• perf.spls regression should be a tune.spls regression?

• network on a block.spls 'solved' but needs to be checked

• examples in all mint.block functions and check all class in \value{} in those functions

• proper examples for mint.pls and mint.spls must be given (I removed the ones currently in as they include Y discrete outcome). See end of my script attached furtherTestingJune2016

• most of the mint.block and block. .Rd files are repetitive. Wonder if we could use some alias to avoid copy paste. Same for tune?

• do we need to import parallel?

• potentially some overlap between tune and tune.multilevel. .Rd files and R function to check and clean up

• MINT.pca

Misc:

• block.pls may need a ridge penalty
• canonical, invariant etc mode for a block.plsda / block.splsda.mint.block.splsda and mint.block.plsda?

[Florian Rohart]

• tune. functions (e.g. mint),tune.mint$choice.keepX should output the keepX number, as tune.mint$selectedvar should output what is current there (website to then be updated)

tune.mint$choice.keepX.constraint gives which variables were selected while tune.splsda$choice.keepX gives the keepX number. Reason is we have two kinds of tuning process: either the variables or the number of variables

[Florian Rohart]

We need to find a better way to report a list of issues like that, cause this is the worst to follow up

[Florian Rohart]

plus: reusing an issue that was a long list of fixed issues make us lose track of all the solved issues. I'd suggest an issue per issue instead of ever-changing lists.