Passing CRAN and expanding test coverage #39
Conversation
Catchup to upstream
clean up weighted sd
| @@ -104,23 +105,20 @@ caretEnsemble <- function(all.models, optFUN=NULL, ...){ | |||
| predict.caretEnsemble <- function(object, keepNA = TRUE, se = NULL, ...){ | |||
There was a problem hiding this comment.
I'm getting this note when I run check():
checking Rd \usage sections ... NOTE
S3 methods shown with full name in documentation object 'predict.caretEnsemble':
‘predict.caretEnsemble’
S3 methods shown with full name in documentation object 'predict.caretStack':
‘predict.caretStack’
The \usage entries for S3 methods should use the \method markup and not
their full name.
See the chapter ‘Writing R documentation files’ in the ‘Writing R
Extensions’ manual.
I think adding the @method tag to predict for caretEnsemble and caretStack fixes this:
There was a problem hiding this comment.
Ah yes, that makes more sense. I will make the change.
There was a problem hiding this comment.
I'm not sure why this happened under Roxygen and why I didn't see it on my machine. I"ll investigate.
There was a problem hiding this comment.
I think this is related to the S4/S3 issue we're discussing elsewhere.
|
|
||
| #' Extract a model accuracy metric from a \code{\link{train}} object. | ||
| #' @rdname metrics | ||
| #' @export |
There was a problem hiding this comment.
I think this should also include:
@method getMetric train
Close #37. Close #38. Accomplish many of the tasks on #32.
Add unit tests for all helper functions for
caretEnsemble. Expand unit test coverage to smaller helper functions like weighted standard deviations. Fixsummary.caretEnsembleto produce correct results for ensembled models.Still need to make the
predict.caretEnsembleoutput more consistent across different options.Fixes a number of issues related to
varImp, though it does not fail gracefully in the case whenvarImpmethod fails for a specific method passed totrain.Added a new test suite using more diverse
trainmethods and some missingness. This checks against some errors relating to building to specifically to the included test models and only increases the size of the package modestly.