User's Guide to ARSPIQ: The Univariate Time Series Analysis Program for Autoregressive Spectral Information Quantile Identification

摘要

The ARSPIQ (AutoRegressive Spectral Information Quantile identification) program is a modified version of ARSPID, a univariate time series program in the TIMESBOARD Computing Library. ARSPIQ provides various diagnostics in the time and frequency domains to help one determine whether a time series is long, short, or no memory. ARSPIQ is written in FORTRAN and consists of a main program and 58 subprograms, many of which are contained in the TIMESBOARD Computing Library. Many of the subprograms used in the quantile analysis are modified version of those used by the ONESAM program. The basic goal of ARSPIQ is to provide diagnostics to aid in modeling a univariate time series. ARSPIQ is not intended to be a modeling or forecasting program, however. The models produced by ARSPIQ are intended only as guides to formulate more complete, rigorous, or valid models. The author's approach is to run ARSPIQ to obtain useful model building diagnostics and then to employ ARSPID in the more formal model building stage, i.e., estimating parameters and checking residuals. With this goal in mind, ARSPIQ has been made as fully automatic as possible with logical defaults provided when input options are specified to be zero.