jstriaukas / midasml

midasml package is dedicated to run predictive high-dimensional mixed data sampling models

midasml

midasml - Estimation and Prediction Methods for High-Dimensional Mixed Frequency Time Series and Panel Data

About

The midasml package implements estimation and prediction methods for high-dimensional mixed-frequency (MIDAS) time-series and panel data regression models. The regularized MIDAS models are estimated using orthogonal (e.g. Legendre) polynomials and sparse-group LASSO estimator. For more information on the midasml approach see ¹²³.

The package is equipped with the fast implementation of the sparse-group LASSO estimator by means of proximal block coordinate descent. High-dimensional mixed frequency time-series data can also be easily manipulated with functions provided in the package.

Software in other languages

• Julia implmentation of the midasml method is available here.
• MATLAB implmentation of the midasml method is available here.
• Python implmentation of the midasml method is being developed at here.

Run to install the package

# CRAN version - 0.1.10
install.packages("midasml") 

# Development version - 0.1.10
# install.packages("devtools")
library(devtools)
install_github("jstriaukas/midasml")

Acknowledgements

Jonas Striaukas acknowledges that this material is based upon work supported by the Fund for Scientific Research-FNRS (Belgian National Fund for Scientific Research) under Grant #FC21388.

References

Footnotes

1. Babii, A., Ghysels, E., & Striaukas, J. Machine learning time series regressions with an application to nowcasting, (2022) Journal of Business & Economic Statistics, Volume 40, Issue 3, 1094-1106. https://doi.org/10.1080/07350015.2021.1899933. ↩

2. Babii, A., Ghysels, E., & Striaukas, J. High-dimensional Granger causality tests with an application to VIX and news, (2022) Journal of Financial Econometrics, Forthcoming. ↩

3. Babii, A., R. Ball, Ghysels, E., & Striaukas, J. Machine learning panel data regressions with heavy-tailed dependent data: Theory and application, (2022) Journal of Econometrics, Forthcoming. ↩

midasml's Issues

sglfitF convergence

Fix: prox_sgl - stop after a certain number of iterations, return the non-convergence flag.
comment: if the min problem is non-converging (e.g. data irregularities) prox_sgl will not escape the loop.

Fix: sglfit - add data check-ups.
comment: if y,x has NA, Inf, stop. if y,x has large outliers, report a warning.

cv.panel.sglfit arguments setting

The explanation of "gindex" is p by 1 vector indicating group membership of each covariate. But I failed to understand how to set it accurately, could you give me some suggestions?

sglfit standardize issue

standardize = TRUE in sglfit leads to unexpected output.

Thanks to Marie Ternes for spotting this.

NOTE: using sg-MIDAS-LASSO model, covariates should be standardized at the original frequency leading to standardize = FALSE.

Question of MIDAS with panel data

Hello,
I am trying to using the MIDAS model to analyze the panel data. But now I have some problems about it and I am wondering if you could give me some advice.
Does your midasml package has a function for panel data?
If so, Can you share me some relevant codes or examples ？

Midas regression with daily, monthly and quarterly data

Hi Dr. Striaukas,

in my lasso midas regression I try to explain quarterly gdp with monthly and daily data. Now I want apply reg.sgl and I realized that it is not possible to use covariates with different frequencies. Because T in matrix x (in your reg.sgl function) is different for daily and monthly data. Can I add more than one matrix x oder did I have overseen something? Appreciate your help.

Kind Regards

Dynamic forecasting

Hello,

Thank you for your great package! I wonder how I can accomplish the dynamic forecasting: that is, estimates for the target variable at the times it was not observed, for example on the daily level.

Thank you!

monthly data using a daily data

Hi,

can we use the "midaslm"package to predict a monthly data using a daily data?

thank you.

T

midasml Tutorial

Hi,

I am interested in applying this approach to one of my research papers. Sadly, I have not found a tutorial on how to use your package. Is there any example code/project available that I could use to understand better how to apply your R code?

Thank you very much!

Best,
Jonathan

Wald statistic for the high-dimensional Granger causality tests

In paragraph 2.4 of "High-Dimensional Granger Causality Tests with an Application to VIX and News", a null hypothesis is tested. Is the Wald statistic discussed on the same page available in midas-ml?

parallel code for cv.panel.sglfit - typo

a typo in the foreach line:

outlist <- foreach(i = seq(nfolds), .packages = c("midsaml")) %dopar%{

->

outlist <- foreach(i = seq(nfolds), .packages = c("midasml")) %dopar%{