Functions and data are provided that support a course that emphasizes statistical issues of inference and generalizability. The functions are designed to make it straightforward to illustrate the use of cross-validation, the training/test approach, simulation, and model-based estimates of accuracy. Methods considered are Generalized Additive Modeling, Linear and Quadratic Discriminant Analysis, Tree-based methods, and Random Forests.
Changes in gamclass version 0.57
Changes in vignettes
o Argument reg.line=NA, in calls to the car functions
sp() and spm() has been changed to regLine=TRUE, as
required in car versions >=3.0.0 .
o An argument diag="boxplot", in figs5.Rnw in a call to
car::spm(), becomes diagonal=list(method="boxplot"),
as required in car versions >=3.0.0 .
Changes in gamclass version 0.53
NEW FEATURES o Functions plotFars() an tabFarsDead() do not now have an argument 'data'. Instead, to deal with a "No visible binding" message, data('FARS', package='gamclass', envir=environment()) is used to place the dataset 'FARS' in the function environment, with FARS <- get("FARS", envir=environment()) used to appease the CRAN check. o The functions eventCounts() (counts of events by specified intervals), gamRF() (use repeated sampling to compare interpolation accuracy of GAM model + randomForests fit to residual), addhlines() (plot horizontal lines to show rpart fitted values) and bssBYcut() (between group SS for y, for all possible splits on a value of x) o Vignettes figs7 and figs8 are new. The former vignettes figs7:figs9 have become figs9:figs11 o The dataset airAccs has been added.