Abstract | ||
---|---|---|
In this paper, a data-driven test is proposed to compare two independent or dependent stationary time series, in terms of the second order dynamics. We show that the problem of time series comparison is equivalent to a goodness-of-fit test checking if a constant model is adequate. Using the same framework, the proposed test is easily extended to compare multiple time series and time series of different lengths. Different to previous methods, it is based on generalized score statistics in an estimating equation setting, with some weak and flexible conditions. An extensive simulation study illustrates the validity of the asymptotic result and finite sample properties, using the tapered periodogram. The proposed test is found to perform well for many different situations, including time series with heavy-tailed or skewed innovations. An application to damage detection using vibration data is discussed. |
Year | DOI | Venue |
---|---|---|
2011 | 10.1016/j.csda.2011.01.013 | Computational Statistics & Data Analysis |
Keywords | Field | DocType |
time series,tapered periodogram,multiple time series,data-driven test,dependent stationary time series,time series comparison,asymptotic result,data-driven,heavy-tailed,autocorrelation,vibration data,different length,goodness-of-fit test checking,different situation,proposed test,generalized score,stationary time series,second order,heavy tail,goodness of fit test,estimating equation | Econometrics,Order of integration,Approximation theory,Stochastic process,Stationary process,Statistics,Score,Goodness of fit,Asymptotic analysis,Mathematics,Autocorrelation | Journal |
Volume | Issue | ISSN |
55 | 6 | Computational Statistics and Data Analysis |
Citations | PageRank | References |
0 | 0.34 | 3 |
Authors | ||
1 |