Simulation of vector nonlinear time series models on clusters

Vector functional coefficient autoregressive (VFCAR) models are powerful nonlinear times series models. Statistical estimation and inference involving VFCAR models are computationally intensive so that simulations to study statistical properties take weeks, or even months. As a consequence, empirical results are often based on a relatively small number of simulation replications, sacrificing precision, accuracy and reliability in the interest of "computational expediency". The simulations are amenable for parallelization; however, parallel computing technology has not been implemented in the VFCAR literature. This paper proposes an approach to the parallelization of statistical simulation codes to address long running times, without resorting to extensive code revisions. This approach takes advantage of recent advances in dynamic loop scheduling on workstation clusters to achieve high performance, even with the presence of unpredictable load imbalance factors. Preliminary results indicate that efficiencies in the range 95-99% on 32 processors of a Sun Solaris cluster are achievable.