I have been searching for time-series techniques that can model highly volatile data, but most techniques that are recommended like GARCH, BSTS, TBATS, etc. don't have a very clear-cut way to include external regressors. Are there any alternate popular techniques, especially non-linear or non-parametric ones that use external regressors?
$\begingroup$
$\endgroup$
1
-
$\begingroup$ Neural networks. $\endgroup$– user2974951Commented Jul 18, 2019 at 6:05
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
BSTS has a way to include external regressors. I am actually using this model at the moment and the inclusion is pretty clear in R package bsts.
bsts "This function can be used either with or without contemporaneous predictor variables (in a time series regression)." https://cran.r-project.org/web/packages/bsts/bsts.pdf
Example 7: https://www.rdocumentation.org/packages/bsts/versions/0.9.2/topics/bsts
$\begingroup$
$\endgroup$
You can always model your time series using any non-time-series model you like, including nonlinear or nonparametric regression, then model the time series of residuals using a standard (or other) time series method.
This is the approach taken in R's forecast::auto.arima(), which runs a standard OLS regression with ARIMA errors.
answered Jul 18, 2019 at 6:31