Time series with infinite-order partial copula dependence

Abstract Stationary and ergodic time series can be constructed using an s-vine decomposition based on sets of bivariate copula functions. The extension such processes to infinite sequences is considered shown yield a rich class models that generalizes Gaussian ARMA ARFIMA allow both non-Gaussian marginal behaviour description the serial partial dependence structure. Extensions classical causal invertible representations linear general are proposed investigated. A practical parsimonious method for parameterizing Kendall autocorrelation function developed. potential resulting give improved statistical fits in many applications indicated with example macroeconomic data.