Bogotá - Ciencias - Maestría en Ciencias - Estadística · 2021
Two-regime functional threshold autoregressive model: an empirical approach
It is known in the literature that economic and financial time series, such as stock returns or the exchange rate, present non-linear dynamics like regime switching, time-varying volatility, or volatility clusters, which should be modeled by using the appropriate non-linear models. Given that financial time series are often of high frequency, it is possible to split the series into intervals and treat each interval as a unique functional observational unit. In this work we introduce and explore, via simulation, a two-regime Functional Threshold Autoregressive model of order one, FTAR(2, 1, 1), as an extension of the univariate TAR(1) model, allowing for a discrete regime switching specification in functional time series governed by a scalar threshold process.