Skip to content

Rolling window forecast with rolling demean #633

@gavincyi

Description

@gavincyi

I saw in the documentation that rolling window forecast can be applied with parameter first_obs and last_obs, while I am looking for an approach with minimal runtime overhead to

  1. demean the series in rolling basis, i.e. returns[first_obs:last_obs] - mean(returns[first_obs:last_obs]), and
  2. fit the GARCH model

I wonder if the constant mean is applied on the rolling basis, or actually on the whole timeseries of argument y.

model = arch_model(tseries, vol="GARCH", mean="Constant", ...)
for i in range(len(tseries) - rolling_window):
  model.fit(first_obs=i, last_ob=i + rolling_window - 1, ...)

I tried to look into the source code but could not conclude it in a glance. Could you help address it?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions