Package index • collinear

Multicollinearity Filtering

Remove redundant predictors from modelling datasets. These functions filter variables by pairwise correlation, variance inflation factors, or both, while respecting user-defined predictor priorities.

collinear(): Smart multicollinearity management
collinear_select(): Dual multicollinearity filtering algorithm
cor_select(): Multicollinearity filtering by pairwise correlation threshold
step_collinear() prep(<step_collinear>) bake(<step_collinear>): Tidymodels recipe step for multicollinearity filtering
vif_select(): Multicollinearity filtering by variance inflation factor threshold

Multicollinearity Assessment

Quantify redundancy among predictors. Compute pairwise correlations, variance inflation factors, and summary statistics for datasets with numeric and categorical variables.

collinear_stats(): Compute summary statistics for correlation and VIF
cor_clusters(): Group predictors by hierarchical correlation clustering
cor_cramer(): Quantify association between categorical variables
cor_df(): Compute signed pairwise correlations dataframe
cor_matrix(): Signed pairwise correlation matrix
cor_stats(): Compute summary statistics for absolute pairwise correlations
vif(): Compute variance inflation factors from a correlation matrix
vif_df(): Compute variance inflation factors dataframe
vif_stats(): VIF Statistics

Predictor Ranking

Prioritize predictors for multicollinearity filtering. Rank variables by their association with a response or by their redundancy with other predictors. Supports cross-validation and multiple response types.

f_binomial_gam(): Area under the curve of binomial GAM predictions vs. observations
f_binomial_glm(): Area Under the Curve of Binomial GLM predictions vs. observations
f_binomial_rf(): Area Under the Curve of Binomial Random Forest predictions vs. observations
f_categorical_rf(): Cramer's V of Categorical Random Forest predictions vs. observations
f_count_gam(): R-squared of Poisson GAM predictions vs. observations
f_count_glm(): R-squared of Poisson GLM predictions vs. observations
f_count_rf(): R-squared of Random Forest predictions vs. observations
f_numeric_gam(): R-squared of Gaussian GAM predictions vs. observations
f_numeric_glm(): R-squared of Gaussian GLM predictions vs. observations
f_numeric_rf(): R-squared of Random Forest predictions vs. observations
preference_order(): Rank predictors by importance or multicollinearity
f_auto(): Automatic selection of predictor scoring method
f_auto_rules(): Decision rules for f_auto()
f_functions(): List predictor scoring functions

Target Encoding

Convert categorical predictors to numeric using response values. Implements mean, leave-one-out, and rank encoding methods for seamless integration of categorical variables in correlation and VIF analyses.

target_encoding_lab(): Convert categorical predictors to numeric via target encoding
target_encoding_loo() target_encoding_mean() target_encoding_rank(): Encode categories as response means

Example Data

Sample datasets for exploring package functionality. Includes dataframes with numeric, categorical, and mixed predictor types, plus multiple response encodings.

toy: Toy dataframe with varying levels of multicollinearity
vi: Large example dataframe
vi_predictors: Vector of all predictor names in vi and vi_smol
vi_predictors_categorical: Vector of categorical predictors in vi and vi_smol
vi_predictors_numeric: Vector of numeric predictor names in vi and vi_smol
vi_responses: Vector of response names in vi and vi_smol
vi_smol: Small example dataframe

Validation Experiments

Results from simulation studies used to calibrate adaptive thresholds and validate the equivalence between correlation and VIF filtering.

experiment_adaptive_thresholds: Dataframe resulting from experiment to test the automatic selection of multicollinearity thresholds
experiment_cor_vs_vif: Dataframe with results of experiment comparing correlation and VIF thresholds
gam_cor_to_vif: GAM describing the relationship between correlation and VIF thresholds
prediction_cor_to_vif: Prediction of the model gam_cor_to_vif across correlation values

Print and Summary Methods

S3 methods for displaying and summarizing results from collinear() and related functions.

print(<collinear_output>): Print all collinear selection results of collinear()
print(<collinear_selection>): Print single selection results from collinear
summary(<collinear_output>): Summarize all results of collinear()
summary(<collinear_selection>): Summarize single response selection results of collinear

Variable Type Detection

Identify and classify variables by type. Detect numeric, categorical, logical, and near-zero variance columns in modelling datasets.

identify_categorical_variables(): Find valid categorical variables in a dataframe
identify_logical_variables(): Find logical variables in a dataframe
identify_numeric_variables(): Find valid numeric variables in a dataframe
identify_response_type(): Detect response variable type for model selection
identify_valid_variables(): Find valid numeric, categorical, and logical variables in a dataframe
identify_zero_variance_variables(): Find near-zero variance variables in a dataframe

Modelling Utilities

Helper functions for model fitting and evaluation. Generate formulas, compute performance metrics, and create class-balancing weights.

case_weights(): Generate sample weights for imbalanced responses
model_formula(): Build model formulas from response and predictors
score_auc(): Compute area under the ROC curve between binomial observations and probabilistic predictions
score_cramer(): Compute Cramer's V between categorical observations and predictions
score_r2(): Compute R-squared between numeric observations and predictions

Input Validation

Internal functions for checking and preparing function arguments. Ensure data frames, variable names, and parameters meet requirements.

drop_geometry_column(): Removes geometry Column From sf Dataframes
validate_arg_df(): Check and prepare argument df
validate_arg_df_not_null(): Ensure that argument df is not NULL
validate_arg_encoding_method(): Check and validate argument encoding_method
validate_arg_f(): Check and validate argument f
validate_arg_function_name(): Build hierarchical function names for messages
validate_arg_max_cor(): Check and constrain argument max_cor
validate_arg_max_vif(): Check and constrain argument max_vif
validate_arg_predictors(): Check and validate argument predictors
validate_arg_preference_order(): Check and complete argument preference_order
validate_arg_quiet(): Check and validate argument quiet
validate_arg_responses(): Check and validate arguments response and responses