Enhance documentation for LocalModel

[mirka-henninger]

The range and standard deviation of the features has a huge influence on the distance measure (distance between the data and the interest of interest) and therewith lime results when the LocalModel function is used. This is also discussed, e.g., in this stackoverflow thread for clustering analyses.

In the lime package by Pedersen and Bernesty, features are scaled before the distance measure is computed. However, there might be reasons where scaling is not desired, e.g., when features have some kind of natural scale. But I think it would at least be a good idea to inform the iml user that scaling the features might be advisable when they want to use lime as an interpretation technique as scaling can influence the computed distance. Therefore, I added this text to the documentation of the LocalModel function, so users can decide themselves whether they would like to scale their features or get more information about the topic.

Thanks for providing and taking care of this package!
Mirka

[codecov-commenter]

Codecov Report

Merging #187 (7ba5dd6) into main (cf7266d) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #187   +/-   ##
=======================================
  Coverage   90.04%   90.04%           
=======================================
  Files          17       17           
  Lines        1667     1667           
=======================================
  Hits         1501     1501           
  Misses        166      166           

Impacted Files	Coverage Δ
R/LocalModel.R	97.59% <ø> (ø)

📣 Codecov can now indicate which changes are the most critical in Pull Requests. Learn more

[christophM]

Thanks for the PR.