November 12, 2019
On November 15th. MSSc. Juho Nyholm will defend his doctoral thesis “Essays on noninvertible ARMA models”. His thesis contributes to the time series econometric literature by expanding novel theoretical results in ARMA modeling under noninvertibility and non-Gaussianity.
Nyholm shows how assuming noninvertibility in classical ARMA models allows for modelling more complex and nonlinear patterns in financial time series than conventional causal and invertible ARMA models. In particular, unlike their invertible counterparts, these models are capable of controlling for heteroscedasticity, which is for example, often encountered in stock market returns. In his thesis, Nyholm constructs methods for testing the adequacy of an estimated noninvertible ARMA model. He also provides asymptotic estimation theory for the parameters of the noninvertible ARMA model designed for time series data that exhibits extreme jumps. Nyholm illustrates the applicability of the noninvertible ARMA model by showing how they are capable of capturing nonlinear predictability in a number of U.S. stock portfolio returns.
Juho Nyholm currently works as an economist at the Economics Department of the Ministry of Finance.