Adds option 'statistic' to get interquantile interval

[adamryczkowski]

… instead of coniidence interval in error_statistic function that gets called by the ggplot.resamples.

It is my solution to the issue #952

It can be used as e.g.

res<-caret::resamples(mymodel)
ggplot(res,metric="RMSE",main="GBM vs xgboost", statistic = 'IR')

[codecov-io]

Codecov Report

Merging #975 into master will decrease coverage by 13.12%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #975       +/-   ##
==========================================
- Coverage   32.63%   19.5%   -13.13%     
==========================================
  Files          90      90               
  Lines       13017   12114      -903     
==========================================
- Hits         4248    2363     -1885     
- Misses       8769    9751      +982

Impacted Files	Coverage Δ
R/getTrainPerf.R	0% <0%> (-100%)	⬇️
R/diag.R	0% <0%> (-100%)	⬇️
R/updates.R	0% <0%> (-61.77%)	⬇️
R/safs.R	5.74% <0%> (-56.58%)	⬇️
R/gafs.R	8.83% <0%> (-55.97%)	⬇️
R/dummyVar.R	0% <0%> (-48.6%)	⬇️
R/selectByFilter.R	0% <0%> (-46.96%)	⬇️
R/rfe.R	0% <0%> (-45.85%)	⬇️
R/resamples.R	0% <0%> (-17.65%)	⬇️
R/reg_sims.R	0% <0%> (-14.29%)	⬇️
... and 52 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6e3d92d...ca58a5a. Read the comment docs.

[topepo]

Can you please add documentation for the parameter, add a unit test, run R CMD check, and update the NEWS file too?

[adamryczkowski]

OK. I will look at that in my spare time.

[topepo]

Getting ready to merge this. Can you write/update units tests related to this (and update the NEWS file)? The PR fails travis