- Add formatting to the X and Y axis of the summary plot, also change the dash to an actual negative sign.
This is the first release of mSHAP on CRAN! It is an implementation of the mSHAP algorithm as described by Matthews and Hartman (2021) (See Paper Here), which allows for the computation of SHAP values on two-part models. This package also includes useful functions for visualizing mSHAP and SHAP values. See the README file on the github page for more examples and use cases.
- Some changes to paper code that is more accurate
- Limit the number of variables that are plotted in the plotting functions
- Add data (.buildignore'd) and code pertaining to mshap paper
- Added examples to all function documentation.
- Added capability of observation plot to have character values in the
variable_values
data frame. - Changed
{ggplot2}
,{ggbeeswarm}
and{tidyr}
to imports, instead of suggests. - Added tests for all functions (close to 100% codecov).
- Added documentation to all functions
- Added a vignette called
mshap_plots
that documents different ways to customize and use the plots. - Added parameters
colorscale
,legend.position
,font_family
, andtitle
to thesummary_plot()
function, for greater ease of customization. - Added parameters
fill_colors
,connect_colors
,expected_color
,predicted_color
,title
, andfont_family
to theobservation_plot()
function for greater ease of customization. - Added Travis CI pipeline functionality.
- Added a helper function to make the
mshap
function more robust - Finished the vignette
- Made the warning message in
observation_plot()
go away (it was occurring because there was only one tick mark on each axis, so I added an invisible one) - Adding dependencies and other items to ensure that there are no notes/warnings in
R CMD CHECK
- Fixed bug in
mshap()
function where data frames with the same number of columns but different names were not getting flagged to add additional columns even whenshap_*_names
was specified. - Fix bug that was causing the expected values to be off if the list-mshap was passed as the second argument to
mshap()
- Fixed bug in the summary_plot function with where the names were assigned