Keep variable selection results of subsamples and visualize

[gokceneraslan]

trafficstars

I added a new option, keep.subsampling = FALSE to stabsel function to keep binary subsample variable selection matrix, so that users can visualize subsampling results. For that, I added a third option to plot function, type="subsamples" which uses pheatmap package to plot a heatmap. Code looks like this:

> stab.lasso <- stabsel(x = bodyfat[, -2], y = bodyfat[,2],
                         fitfun = lars.lasso, cutoff = 0.75,
                         PFER = 1, keep.subsampling=T)
> plot(stab.lasso, type='subsamples')

and output is as follows:

Does it make sense to provide a way of visualizing subsample selection results like this?

[hofnerb]

In general I like the idea as it provides new ways to look at the model and at the stability selection results. Things I don't like that much:

• [ ] Currently, the documentation of new argument keep.subsampling and new option type = "subsamples" is missing.

• [ ] The package requires another dependency. Is it possible to realize the plot without pheatmap? If not, please fix:

  checking dependencies in R code ... WARNING
  '::' or ':::' import not declared from: 'pheatmap'
  'loadNamespace' or 'requireNamespace' call not declared from: 'pheatmap'
  

• [ ] The current legend is not really appropriate. Is it possible to provide a binary legend (perhaps also with names "selected" / "not selected") instead?

[hofnerb]

@gokceneraslan (when) will you be able to have a look at the remaining issues?

[gokceneraslan]

I'll have time next week to look into this.