Academic
Publications
Estimating common trends in multivariate time series using dynamic factor analysis

Estimating common trends in multivariate time series using dynamic factor analysis,10.1002/env.611,Environmetrics,A. F. Zuur,R. J. Fryer,I. T. Jolliff

Estimating common trends in multivariate time series using dynamic factor analysis   (Citations: 40)
BibTex | RIS | RefWorks Download
Journal: Environmetrics , vol. 14, no. 7, pp. 665-685, 2003
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...To identify which teleconnection indices might be related to the anchovy/sardine complex, a DFA model (Molenaar, 1985; Zuur et al., 2003a, b; Huang et al., 2006) was developed using the time series of the ratio of the catches in each area as response variables, fitting one common trend and using the teleconnection indices and fishing effort as explanatory variables...

    Isidora KataraGrahamet al. Environmental drivers of the anchovy/sardine complex in the Eastern Me...

    • ...Popular time series techniques such as spectral analysis, wavelet analysis, ARIMA and Box-Jenkins models require stationary time series without missing values, and are not particularly suitable for inferring inter-relationships between the data [1]...
    • ...Recent advances in computing technologies have encouraged the application of more computationally intensive techniques, such as factor analysis for characterizing multiple time series [1], [2],[3]...
    • ...In [1] and [3] the authors propose parametric non-Bayesian approaches and use Akaike’s information criterion (AIC) for model-selection, whereas [2] proposes a parametric Bayesian approach with a fixed number of factors and MCMC based model inference...

    Priyadip Rayet al. Nonparametric Bayesian factor analysis of multiple time series

    • ...Popular multivariate time series analysis techniques such as spectral methods, wavelet analysis and Box-Jenkins models require stationary time series without missing data, and are not particularly suitable for infering inter-relationships across data [1]...
    • ...As seen from (13) and (18), the primary difference between a stick-breaking prior and a kernel stick-breaking prior is that the stick weights are modulated by an additional bounded kernel K(r;r � ,�l) → [0,1] which is a function of the spatial location r. Thus if two cities are spatially proximate, they will have similar stick weights wj(r) and hence are likely to share the same GP...

    Priyadip Rayet al. Non-parametric Bayesian modeling and fusion of spatio-temporal informa...

    • ...With DFA, underlying temporal variation in observed data (response variables) is modeled as linear combinations of common trends (unexplained variability), a constant level parameter, zero or more explanatory variables (additional observed time series), and noise [Zuur et al., 2003b]...
    • ...Additionally, Akaike’s information criterion, AIC [Akaike, 1974] is often used as a decision tool for choosing between competing models [Zuur et al., 2003b]...
    • ...3.2. Dynamic Factor Analysis [14] DFA is based on structural time series models [Harvey, 1989] and aims to describe a set of N time series (termed response variables) using a dynamic factor model (DFM) that includes M common trends (M < N), K explanatory variables, a level or intercept parameter, and noise [Lütkepohl, 1991; Zuur et al., 2003b]:...
    • ...The Kalman filter and smoothing algorithm then skips these missing observations [Zuur et al., 2003b]...
    • ...This allowed us to compare the relative importance of explanatory variables across the set of response variables [Zuur et al., 2003b; Zuur and Pierce, 2004]...

    D. Kaplanet al. Untangling complex shallow groundwater dynamics in the floodplain wetl...

Sort by: