Skip to content
All library documents

Handling Weekend Gaps in GARCH Models of Bitcoin and Fiat Returns

Article Quant Q&A · Author: Mysterio
Read the original

Summary

The document considers how to align daily Bitcoin returns, which include weekends, with fiat currency data that may be missing on those days when fitting a GARCH model. The software package mentioned rejects null observations, leaving the analyst to consider removing Bitcoin weekend rows or filling the missing fiat data.

The accepted response rejects inserting null returns, saying this changes the estimation substantially, and favors dropping weekend rows from the Bitcoin series. This offers a practical alignment choice, but gives no statistical derivation, comparison of alternatives, or empirical results. The guidance is brief and specific to the stated data setup; the document does not discuss other ways of modeling different trading calendars or the consequences of excluding weekend crypto movements.

Key ideas

  • A GARCH input series needs a consistent treatment of dates across assets.
  • The response warns that inserting null returns can substantially alter estimation.
  • Dropping Bitcoin weekend rows is suggested for alignment with fiat data.
  • The document gives no detailed statistical argument or empirical comparison.

Tags

CryptoForexVolatilityStatistics
Full text
# Drop weekend data Vs fill weekend data for GARCH-type modelling


# Drop weekend data Vs fill weekend data for GARCH-type modelling












I have a dilemma for an analysis I'm currently on. I doing some GARCH modelling of bitcoin and a fiat currency.

There are some null values with the fiat datasets in comparison with bitcoin data as cryptos trade on weekends. The R package I'm using `rugarch` won't accept null values

Now, I believe I have two choices:

- Drop weekend rows of bitcoin

- Fill weekend rows of the fiat currency

My question are there statistical justifications for either option and which option hurts least the integrity of my model?

## Answer by DomingoBrown (score 2, accepted)

https://quant.stackexchange.com/a/39609

The second option is not viable. Introducing null returns in your GARCH completly alters your estimation

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.