LasForecast

This package develops a framework for economic forecasting in a data-rich environment with a particular emphasis on linear predictive regression. The goal is to automate the processes of parameter tuning, rolling window forecasting and backtesting, valid inference on parameters of interests, and visualization in a unified framework.

The package covers the following methods.

	Functions	Reference
Lasso	train_lasso(), glmnet::glmnet()	Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B (Methodological), 58(1), 267-288.
Adaptive Lasso (Alasso)	train_lasso(ada = TRUE), adalasso()	Zou, H. (2006). The adaptive lasso and its oracle properties. Journal of the American statistical association, 101(476), 1418-1429 Medeiros, M. C., & Mendes, E. F. (2016). ℓ1-regularization of high-dimensional time-series models with non-Gaussian and heteroskedastic errors. Journal of Econometrics, 191(1), 255-271.
Twin Adaptive Lasso (TAlasso)	train_replasso(ada = TRUE), replasso()	Lee, J. H., Shi, Z., & Gao, Z. (2022). On LASSO for predictive regression. Journal of Econometrics, 229(2), 322-349.
Best Subset Selection	bss.bic(), bss()	Bertsimas, D., King, A., & Mazumder, R. (2016). Best subset selection via a modern optimization lens. The annals of statistics, 44(2), 813-852.
Complete Subset Regression (CSR)	csr.bic(), csr()	Elliott, G., Gargano, A., & Timmermann, A. (2013). Complete subset regressions. Journal of Econometrics, 177(2), 357-373. Elliott, G., Gargano, A., & Timmermann, A. (2015). Complete subset regressions with large-dimensional sets of predictors. Journal of Economic Dynamics and Control, 54, 86-110.
L2-relaxation Forecast Combination	l2relax(), train_l2_relax()	Shi, Z., Su, L., & Xie, T. (2024). : $\ell_2$-Relaxation: With Applications to Forecast Combination and Portfolio Analysis, The Review of Economics and Statistics
XDlasso: IVX-Desparsified Lasso	debias_ivx()	Gao, Z., Lee, J.H., Mei, Z.W., & Shi, Z. (2024). On LASSO Inference for High Dimensional Predictive Regression.

The package exploits the advantage of well-established packages like glmnet and model training framework caret.

We can run backtesting and compare the forecasting performance among different methods based on rolling windows and forecasting horizons using the roll_predict function. For example, we can replicate the empirical results in Lee, Shi and Gao (2022) as in this script.

Installation

install.packages("devtools")
devtools::install_github("zhan-gao/LasForecast")