Package index • smimodel

Estimating SMI model

Functions to estimate a SMI model with or without penalty parameter tuning.

model_smimodel(): Sparse Multiple Index (SMI) Models

greedy_smimodel(): SMI model estimation through a greedy search for penalty parameters

Estimating other benchmark models

Functions to fit some benchmark comparison methods.

model_backward(): Nonparametric Additive Model with Backward Elimination

model_gaim(): Groupwise Additive Index Models (GAIM)

model_ppr(): Projection Pursuit Regression (PPR) models

model_gam(): Generalised Additive Models

model_lm(): Linear Regression models

Augment methods

Obtain residuals and fitted values of the fitted models.

augment(<smimodel>): Augment function for class smimodel

augment(<backward>): Augment function for class backward

augment(<gaimFit>): Augment function for class gaimFit

augment(<pprFit>): Augment function for class pprFit

augment(<gamFit>): Augment function for class gamFit

augment(<lmFit>): Augment function for class lmFit

Predict methods

Obtain residuals and fitted values of the fitted models.

predict(<smimodel>): Obtaining forecasts on a test set from a fitted smimodel

predict(<backward>): Obtaining forecasts on a test set from a fitted backward

predict(<gaimFit>): Obtaining forecasts on a test set from a fitted gaimFit

predict(<pprFit>): Obtaining forecasts on a test set from a fitted pprFit

predict(<gamFit>): Obtaining forecasts on a test set from a fitted gamFit

predict(<lmFit>): Obtaining forecasts on a test set from a fitted lmFit

predict(<cgaim>): Predictions from a fitted CGAIM object - copied from cgaim:::predict.cgaim() and modified

Point estimate accuracy measures

Calculate point estimate accuracy measures.

MSE() MAE() point_measures: Point estimate accuracy measures

Prediction Intervals

Functions to construct and evaluate prediction intervals in time series forecasting problems.

bb_cvforecast(): Single season block bootstrap prediction intervals through time series cross-validation forecasting

cb_cvforecast(): Conformal bootstrap prediction intervals through time series cross-validation forecasting

avgCoverage(): Calculate interval forecast coverage

avgWidth(): Calculate interval forecast width

Other user-facing functions

Other exported functions that can be called by users.

autoplot(<smimodel>): Plot estimated smooths from a fitted smimodel

residuals(<smimodel>): Extract residuals from a fitted smimodel

lag_matrix(): Function for adding lags of time series variables

print(<smimodel>): Printing a smimodel object

print(<smimodelFit>): Printing a smimodelFit object

print(<backward>): Printing a backward object

print(<gaimFit>): Printing a gaimFit object

print(<pprFit>): Printing a pprFit object

forecast(<smimodel>): Forecasting using SMI models

forecast(<backward>): Forecasting using nonparametric additive models with backward elimination

forecast(<gaimFit>): Forecasting using GAIMs

forecast(<pprFit>): Forecasting using PPR models

forecast(<gamFit>): Forecasting using GAMs

leadlagMat(): Create lags or leads of a matrix

Other non-user-facing functions

Other internal functions that are not exported.

smimodel.fit(): SMI model estimation

new_smimodelFit(): Constructor function for the class smimodelFit

update_smimodelFit(): Updating a smimodelFit

make_smimodelFit(): Converting a fitted gam object to a smimodelFit object

inner_update(): Updating index coefficients and non-linear functions iteratively

init_alpha(): Initialising index coefficients

update_alpha(): Updating index coefficients using MIP

greedy.fit(): Greedy search for tuning penalty parameters

tune_smimodel(): SMI model with a given penalty parameter combination

normalise_alpha(): Scaling index coefficient vectors to have unit norm

loss(): Calculating the loss of the MIP used to estimate a SMI model

allpred_index(): Constructing index coefficient vectors with all predictors in each index

split_index(): Splitting predictors into multiple indices

scaling(): Scale data

unscaling(): Unscale a fitted smimodel

augment(<smimodelFit>): Augment function for class smimodelFit

predict(<smimodelFit>): Obtaining forecasts on a test set from a smimodelFit

predict_gam(): Obtaining recursive forecasts on a test set from a fitted mgcv::gam

eliminate(): Eliminate a variable and fit a nonparametric additive model

blockBootstrap(): Futures through single season block bootstrapping

residBootstrap(): Generate multiple single season block bootstrap series

seasonBootstrap(): Single season block bootstrap

randomBlock(): Randomly sampling a block

possibleFutures_smimodel(): Possible future sample paths (multi-step) from smimodel residuals

possibleFutures_benchmark(): Possible future sample paths (multi-step) from residuals of a fitted benchmark model

prep_newdata(): Prepare a data set for recursive forecasting

remove_lags(): Remove actual values from a data set for recursive forecasting

truncate_vars(): Truncating predictors to be in the in-sample range

Package metadata

Package information.

smimodel smimodel-package: smimodel: Sparse Multiple Index (SMI) Models for High-dimensional Nonparametric Forecasting