An Interface to Statistical Modeling Independent of Model Architecture
Plot out model values
mosaicModel package
Function builder for proportions.
Compute sensible values from a data set for use as a baseline
Function builder for confidence intervals on proportions
Calculate measures of collinearity
Construct a call for refitting a model from the model itself
Extract training data from model
Formula interface to counts
Joint and conditional proportions
Find typical levels of explanatory variables in a model/dataset.
Get the names of the explanatory or response variables in a model
Extract the model formula used in specifying the model
Interval statistics for use with df_stats()
Compare models with k-fold cross validation
Calculate effect sizes in a model
Create bootstrapped ensembles of a model
Mean square prediction error
Evaluate a model for specified inputs
Internal functions for evaluating models
Transforms a model into a function of inputs -> output
Provides functions for evaluating, displaying, and interpreting statistical models. The goal is to abstract the operations on models from the particular architecture of the model. For instance, calculating effect sizes rather than looking at coefficients. The package includes interfaces to both regression and classification architectures, including lm(), glm(), rlm() in 'MASS', random forests and recursive partitioning, k-nearest neighbors, linear and quadratic discriminant analysis, and models produced by the 'caret' package's train(). It's straightforward to add in other other model architectures.