This thesis focuses on the application of the Generalized Autoregressive Score (GAS) model to the multivariate time series analysis of returns from four major cryptocurrencies: Bitcoin, Ethereum, Litecoin, and Ripple, over the period from early 2016 to the end of 2021. The model is evaluated both in terms of its in sample fit to the data but also for its out of sample forecasting performance (point, distributional, and interval forecasts). The results are then compared with those obtained from the more traditional DCC-GARCH (Dynamic Conditional Correlation GARCH) model. The conclusion is that the GAS model outperforms traditional alternatives, thanks to its score-driven parameter updating mechanism which makes it more robust to outliers. Using various statistical tools, it is also shown that the performance differences between the GAS and DCC-GARCH models are statistically significant.
La presente relazione si concentra sull’applicazione del modello GAS (Generalized Autoregressive Score) all’analisi della serie storica multivariata dei rendimenti di quattro delle principali criptovalute: Bitcoin, Ethereum, Litecoin e Ripple, nel periodo compreso tra l’inizio del 2016 e la fine del 2021. Del modello si valuta non solo l'adattamento ai dati, ma anche le performance previsive (puntuali, distributive e intervallari) e i risultati vengono messi a confronto con quelli ottenuti dal più tradizionale modello DCC-GARCH (Dynamic Conditional Correlation GARCH). Si giunge alla conclusione che il modello GAS, sia più performante delle alternative tradizionali in quanto, grazie al meccanismo di aggiornamento dei parametri score-driven, esso risulta essere più robusto alla presenza di valori estremi. Utilizzando diversi strumenti statistici si è poi in grado di evidenziare come le differenze di performance fra il GAS e il DCC GARCH siano significative.
Modello GAS per la volatilità delle principali criptovalute
TESTA, LEONARDO
2024/2025
Abstract
This thesis focuses on the application of the Generalized Autoregressive Score (GAS) model to the multivariate time series analysis of returns from four major cryptocurrencies: Bitcoin, Ethereum, Litecoin, and Ripple, over the period from early 2016 to the end of 2021. The model is evaluated both in terms of its in sample fit to the data but also for its out of sample forecasting performance (point, distributional, and interval forecasts). The results are then compared with those obtained from the more traditional DCC-GARCH (Dynamic Conditional Correlation GARCH) model. The conclusion is that the GAS model outperforms traditional alternatives, thanks to its score-driven parameter updating mechanism which makes it more robust to outliers. Using various statistical tools, it is also shown that the performance differences between the GAS and DCC-GARCH models are statistically significant.| File | Dimensione | Formato | |
|---|---|---|---|
|
Testa_Leonardo.pdf
accesso aperto
Dimensione
5.01 MB
Formato
Adobe PDF
|
5.01 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/92982