Bayesian Dynamic Modelling: Bayesian Theory and Applications
Bayesian Dynamic Modelling: Bayesian Theory and Applications
Bayesian Dynamic Modelling: Bayesian Theory and Applications
Mike West
Department of Statistical Science
Duke University
Chapter 8 of
Bayesian Theory and Applications
published in honour of Professor Sir Adrian F.M. Smith
(eds: P. Damien and P. Dellaportas and N. G. Polson and D. A. Stephens)
Clarendon: Oxford University Press, 2013
1.1 Introduction
Bayesian time series and forecasting is a very broad field and any attempt at
other than a very selective and personal overview of core and recent areas would
be foolhardy. This chapter therefore selectively notes some key models and ideas,
leavened with extracts from a few time series analysis and forecasting examples.
For definitive development of core theory and methodology of Bayesian state-
space models, readers are referred to [74,46] and might usefully read this chap-
ter with one or both of the texts at hand for delving much further and deeper.
The latter parts of the chapter link into and discuss a range of recent develop-
ments on specific modelling and applied topics in exciting and challenging areas
of Bayesian time series analysis.
If the φtj are constant over time, xt is a standard AR(p) process; in this sense,
the main class of traditional linear time series models is a special case of the class
of DLMs.
F 0 = (1, ft , E20 ,
E20 , E20 ),
φ1 φ2 (1.2)
G = block diag 1, 1, H(a1 ), H(a2 ),
1 0
where:
• θt1 is a local level/random walk intercept varying in time;
• θt2 is a dynamic regression parameter in the regression on the univariate
predictor/independent variable time series ft ;
• ρtj is a seasonal/periodic component of wavelength 2π/aj for j = 1, 2,
with time-varying amplitudes and phases– often an overall annual pattern
in weekly or monthly data, for example, can be represented in terms of
a set of harmonics of the fundamental frequency, such as would arise in
the example here with a1 = π/6, a2 = π/3 yielding an annual cycle and a
semi-annual (six month) cycle;
• zt is an AR(2) process– a short-term correlated underlying latent process–
that represents residual structure in the time series signal not already cap-
tured by the other components.
chapter 10 of [74] and, with recent updates and new extensions, in chapters 4,7
and 10 of [46].
The use of discount factors to structure evolution variance matrices has been
and remains central to many applications (chapter 6 of [74]). In models with
non-trivial state vector dimension p, we must maintain control over specification
of the Wt to avoid exploding the numbers of free parameters. In many cases, we
are using Wt to reflect low levels of stochastic change in elements of the state.
When the model is structured in terms of block components via superposition
as described above, the Wt matrix is naturally structured in a corresponding
block diagonal form; then the strategy of specifying these blocks in Wt using
the discount factor approach is natural (section 4.3.6 of [46]). This strategy
describes the innovations for each component of the state vector as contributing
a constant stochastic “rate of loss of information” per time point, and these rates
may be chosen as different for different components. In our example above, a
dynamic regression parameter might be expected to vary less rapidly over time
than, perhaps, the underlying local trend.
Central to many applications of Bayesian forecasting, especially in commer-
cial and economic studies, is the role of “open modelling”. That is, a model is
often one of multiple ways of describing a problem, and as such should be open
to modification over time as well as integration with other formal descriptions
of a forecasting problem (chapter 1 of [74]). The role of statistical theory in
guiding changes– interventions to adapt a model at any given time based on
additional information– that maintain consistency with the model is then key.
Formal sequential analysis in a DLM framework can often manage this via ap-
propriate changes in the variance components. For example, treating a single
observation as of poorer quality, or a likely outlier, can be done via an inflated
variance multiplier kt ; feeding into the model new/external information that
suggests increased chances of more abrupt change in one or more components
of a state vector can be done via larger values of the corresponding elements
of Wt , typically using a lower discount factor in the specification for just that
time, or times, when larger changes are anticipated. Detailed development of a
range of subjective monitoring and model adaptation methods of these forms,
with examples, are given in chapters 10-12 of [74] and throughout [43]; see
also chapter 4 of [46] and earlier relevant papers [72, 62, 73].
∗ r
where each ztj is an underlying latent process: each ztj is a TVAR(1) process
c
and each ztj is a quasi-cyclical time-varying process whose characteristics are
effectively those of a TVAR(2) overlaid with low levels of additive noise, and that
exhibits time-varying periodicities with stochastically varying amplitude, phase
and period. In the special case of constant AR parameters, the periods of these
c
quasi-cyclical ztj processes are also constant.
This DLM decomposition theory underlies the use of these models– state-
space models/DLMs with AR and TVAR components– for problems in which
we are interested in a potentially very complicated and dynamic autocorrela-
tion structure, and aim to explore underlying contributions to the overall signal
that may exhibit periodicities of a time-varying nature. Many examples appear
in [74, 46] and references there as well as the core papers referenced above.
Figures 1.1, 1.2 and 1.3 exemplify some aspects of this in the analysis of the
climatological Southern Oscillation Index (SOI) time series of Section 1.3.2.
Also, write φ = (φ1 , φ2 )0 for the AR parameters of the latent AR(2) model com-
ponent. The DLM can be fitted using standard theory assuming the full set of
model parameters µ = {v, φ, w, τ1:4 } to be known. Given these parameters, the
forward-filtering and smoothing based on normal/linear theory applies.
Markov chain Monte Carlo methods naturally open the path to a complete
Bayesian analysis under any specified prior p(µ); see chapter 15 of [74] and sec-
tion 4.5 of [46] for full details and copious references, as well as challenging
applications in chapter 7 of [46]. Given an observed data sequence y1:n , MCMC
Computation and Model Enrichment 7
F IG . 1.1. Left frame: Approximate posterior 95% credible intervals for the moduli of the 12 latent
AR roots in the AR component of the model fitted to the SOI time series. Right frame: Approxi-
c with largest wavelength,
mate posterior for the wavelength of the latent process component ztj
indicating a dominant quasi-periodicity in the range of 40-70 months.
Figures 1.1, 1.2 and 1.3 show aspects of an analysis of the climatological
Southern Oscillation Index (SOI) time series. This is a series of 540 monthly
observations computed as the “difference of the departure from the long-term
monthly mean sea level pressures” at Tahiti in the South Pacific and Darwin in
8 Bayesian Dynamic Modelling
Northern Australia. The index is one measure of the so-called “El Nino-Southern
Oscillation”– an event of critical importance and interest in climatological studies
in recent decades and that is generally understood to vary periodically with a
very noisy 3-6 year period of quasi-cyclic pattern. As discussed in [24]– which
also details the history of the data and prior analyses– one of several applied
interests in this data is in improved understanding of these quasi-periodicities
and also potential non-stationary trends, in the context of substantial levels of
observational noise.
The DLM chosen here is yt = θt1 +zt +νt where θt1 is a first-order polynomial
local level/trend and zt is an AR(12) process. The data is monthly data over the
year, so the AR component provides opportunities to identify even quite subtle
longer term (multi-year) periodicities that may show quite high levels of stochas-
tic variation over time in amplitude and phase. Extensions to TVAR components
F IG . 1.2. Upper frame: Scatter plot of the monthly SOI index time series superimposed on the
trajectories of the posterior mean and a few posterior samples of the underlying trend. Lower
frame: SOI time series followed by a single synthetic future– a sample from the posterior pre-
dictive distribution over the three or fours years following the end of the data in 1995; the
corresponding sample of the predicted underlying trend is also shown.
Computation and Model Enrichment 9
F IG . 1.3. Aspects of decomposition analysis of the SOI series. Upper frame: Posterior means of
(from the bottom up) the latent AR(12) component zt (labelled as “data”), followed by the three
extracted component ztjc for j = 1, 2, 3, ordered in terms of decreasing estimated periods; all
are plotted on the same vertical scale, and the AR(12) process is the direct sum of these three
and subsidiary components. Lower frame: A few posterior samples (in grey) of the latent AR(12)
process underlying the SOI series, with the approximate posterior mean superimposed.
would allow the associated periods to also vary as discussed and referenced
above. Here the model parameters include the 12-dimensional AR parameter φ
that can be converted to autoregressive roots (section 9.5 of [74]) to explore
whether the AR component appears to be consistent with an underlying station-
ary process or not as well as to make inferences on the periods/wavelengths of
any identified quasi-periodic components. The analysis also defines posterior in-
r c
ferences for the time trajectories of all latent components ztj and ztj by applying
the decomposition theory to each of the posterior simulation samples of the state
vector sequence θ0:n .
Figures 1.1 shows approximate posteriors for the moduli of the 12 latent AR
roots, all very likely positive and almost surely less than 1, indicating station-
arity of zt in this model description. The figure also shows the corresponding
10 Bayesian Dynamic Modelling
c
posterior for the wavelength of the latent process component ztj having highest
wavelength, indicating a dominant quasi-periodicity in the data with wavelength
between 40-70 months– a noisy “4-year” phenomenon, consistent with expecta-
tions and prior studies. Figure 1.2 shows a few posterior samples of the time
trajectory of the latent trend θt1 together with its approximate posterior mean,
superimposed on the data. The inference is that of very limited change over time
in the trend in the context of other model components. This figure also shows the
data plotted together with a “synthetic future” over the next three years: that is,
a single draw from the posterior predictive distribution into the future. From the
viewpoint of model fit, exploring such synthetic futures via repeat simulations
studied by eye in comparison with the data can be most informative; they also
feed into formal predictive evaluations for excursions away from (above/below)
the mean, for example [24].
Additional aspects of the decomposition analysis are represented by Fig-
ures 1.3. The first frame shows the posterior mean of the fitted AR(12) compo-
nent plotted over time (labelled as “data” in the upper figure), together with the
corresponding posterior mean trajectories of the three latent quasi-cyclical com-
ponents having largest inferred periods, all plotted on the same vertical scale.
Evidently, the dominant period component explains much of the structure in
the AR(12) process, the second contributing much of the additional variation at
a lower wavelength (a few months). The remaining components contribute to
partitioning the noise in the series and have much lower amplitudes. The figure
also shows several posterior draws for the zt processes to give some indication
of the levels of uncertainty about its form over the years.
where N (·, ·, ·) denotes a matrix normal distribution (section 10.6 of [46]). Here
the row vector yt0 follows a DLM with a matrix state Θt . The q × q time-varying
variance matrix Σt determines patterns of co-changes in observation and the
latent matrix state over time. These models are building blocks of larger (factor,
hierarchical) models of increasing use in financial time series and econometrics;
see, for example, [49, 48], chapter 16 of [74] and chapter 10 of [46].
Modelling multivariate stochastic volatility– the evolution over time of the
variance matrix series Σt – is central to these multivariate extensions of DLMs.
The first multivariate stochastic volatility models based on variance matrix dis-
count learning [50, 51], later developed via matrix-beta evolution models [59,
60], remain central to many implementations of Bayesian forecasting in finance.
Here Σt evolves over one time interval via a non-linear stochastic process model
involving a matrix beta random innovation inducing priors and posteriors of
conditional inverse Wishart forms. The conditionally conjugate structure of the
exchangeable model form for {Θt , Σt }, coupled with discount factor-based spec-
ification of the Wt evolution variance matrices, leads to a direct extension of the
closed form sequential learning and retrospective sampling analysis of the uni-
variate case (chapter 10 of [46]). In multiple studies, these models have proven
Multivariate Time Series 13
F IG . 1.4. Daily prices of the Japanese Yen in US dollars over several years in the 1980s-1990s,
followed by plots of forecasts from the multivariate TV-TVAR model with stochastic volatility
fitted to a 12-dimensional FX time series of which the Yen is one element. The shaded forecast
region is made up of 75 sets of 3-day ahead forecasts based on the sequential analysis: on each
day, the “current” posterior for model states and volatility matrices is simulated to generate
forecasts over the next 3 days.
F IG . 1.5. Upper frames: Images of posterior estimates of the 12 × 12 precision matrices Σ−1
t in
the analysis of the multivariate currency prices time series. The differences in patterns visually
evident reflect the extent and nature of changes in the volatility structure across time as rep-
resented by the four selected time points spaced apart by a few hundred days. Lower frame:
Percent variation explained by the first three dominant principal components of the posterior
mean of Σt for each time t = 1 : n over the FX time series, illustrating the nature of variation in
the contribution of the main underlying “common components” of volatility in the 12 currency
price series over the ten year period.
Σt−1 at 4 selected time points capture some flavour of time variation, while the
percent of the total variation in the posterior mean of Σt explained by the first
three dominant principal components at each t captures additional aspects.
1.4.2 Multivariate Normal DLMs: Dynamic Latent Factor and TV-VAR Models
Time-varying vector autoregressive (TV-VAR) models define a rich and flexible
approach to modelling multivariate structure that allows the predictive relation-
ships among individual scalar elements of the time series to evolve over time.
The above section already described the use of such a model within the ex-
changeable time series framework. Another way in which TV-VAR models are
used is to represent the dynamic evolution of a vector of latent factors underly-
Some Recent and Current Developments 15
ing structure in a higher-dimensional data series. One set of such dynamic latent
factor TV-VAR models has the form
yt = F
Pt θt + Bt xt + νt , νt ∼ N (0, Ψ),
(1.5)
xt = i=1:p Ati xt−i + ωt , ωt ∼ N (0, Σt ).
Here xt is a latent k−vector state process following a TV-VAR(p) model with, typ-
ically, k << q so that the common structure among the elements of yt is heavily
driven by a far lower-dimensional dynamic state. The set of p, q × q autore-
gressive coefficient matrices Ati are often time-varying with elements modelled
via, for example, sets of univariate AR(1) processes or random walks. The fac-
tor loadings matrix Bt maps factors to data; in some models, including prior
Bayesian factor analysis approaches [2], this will be taken as constant. The k × k
dynamic covariance matrix Σt drives the innovations of the state evolution, and
allowing for stochastic volatility here defines an enrichment of the TV-VAR struc-
ture. The additional component Ft θt superimposed may include dynamic re-
gression and other terms, with relevant state evolution equations for θt . Such
models are receiving increasing use in natural science and engineering appli-
cations [45, 75, 15, 16, 20, 46] as well as in econometrics and finance [36, 37].
Chapters 8 and 9 of [46] describe aspects of the theory and methodology of vec-
tor AR and TVAR models, connections with latent factor modelling in studies of
multiple time series in the neurosciences, and discussion of multiple other cre-
ative applications of specialized variants of this rich class of models. The model
contains traditional DLMs, VAR models, latent factor models as previously de-
veloped as well as the more elaborate TV-VAR factor forms.
graphical models, coupled with learning about graphical model structures based
on existing model search methods [13, 25]. Applications in financial time se-
ries for predictive portfolio analysis show improvements in portfolio outcomes
that illustrate the practical benefits of the parsimony induced via appropriate
graphical model structuring in multivariate dynamic modelling [10, 9].
These developments have extended to contexts of matrix time series [61]
for applications in econometrics and related areas. Building on Bayesian anal-
yses of matrix-variate normal distributions, conditional independence graphical
structuring of the characterizing variance matrix parameters of such distribu-
tions again opens the path to parsimonious structuring of models for increas-
ingly high-dimensional problems. This is complemented by the development of
a broad class of dynamic models for matrix-variate time series within which
stochastic elements defining time series errors and structural changes over time
are subject to graphical model structuring.
Σt = Ψt + Υt Σt−1 Υ0t
where the q ×q random innovations matrices Υt and Ψt have joint matrix normal,
inverse Wishart distributions independently over time. Conditional means have
the form
E(Σt |Σt−1 ) = S + R(Σt−1 − S)R0 + Ct (Σt−1 )
F IG . 1.7. Examples of latent thresholding from an analysis of a 3-dimensional economic time se-
ries using a TV-VAR(2) model (data from [36]). Upper four frames: Trajectories of approximate
posterior means of 4 of the (18) latent time-varying autoregressive coefficients; the grey shading
indicates estimated time-varying posterior probabilities of zero coefficients from the LTM con-
struction. Lower four images: Images showing estimates of posterior probabilities (white = high,
black = low) of non-zero entries in dynamic precision matrices Σ−1 t modelled using an LTM
extension of the Cholesky AR(1) model [32]. The data are time series on q = 12 daily interna-
tional exchange rates (data from [46]) and the images show posterior sparsity probabilities for
the 12×12 matrices at four selected time points, indicating both the ability of the LTM to identify
zeros as well as how the sparsity pattern changes over time based on latent thresholding.
approach replaces the sequence β1:n with the thresholded version b1:n where
bt = βt I(|βt | < τ ) for each t and based on some threshold τ. The concept is sim-
ple: the coefficient process is relevant, taking non-zero values, only when it beats
the threshold, otherwise it is deemed insignificant and shrunk to zero. Extend-
ing MCMC analyses of multivariate DLMs to integrate the implied hierarchical
model components, now embedding mutliple latent processes underlying the
actual thresholded parameter processes, states and factors, requires substantial
computational development, as detailed in [36]. The payoffs can be meaning-
20 Bayesian Dynamic Modelling
where the j th observation comes at real-time tj and the spacings between con-
secutive observations are typically far greater than the fine time step h. Here xt
represents the underlying state vector of the systems (levels of gene or protein
expression, numbers of cells, etc.), Θ all model parameters, ν∗ measurement er-
ror and ω∗ state evolution noise. The density forms to the right indicate more
general model forms in which errors and noise may not be additive, when only
partial observation is made on the state, and so forth.
The forefront research challenges in this area include development of effi-
cient and effective computations for posterior inference and model comparison.
Increasingly, SMC methods including SMC/importance sampling and ABC-SMC
are being explored and evaluated, with models of increasing dimension and com-
plexity [35, 52, 58, 4, 77]. In coming years, complex, multivariate dynamical
Some Recent and Current Developments 21
systems studies in biology are sure to define a major growth area for Bayesian
dynamic models and time series analysis.
Acknowledgement
The author is grateful to Ioanna Manolopoulou and Emily Fox for comments on
this chapter. This work was partly supported by grant DMS-1106516 from the
U.S. National Science Foundation (NSF), and grants P50-GM081883 and RC1-
AI086032 of the U.S National Institutes of Health. Any opinions, findings and
conclusions or recommendations expressed in this work are those of the author
and do not necessarily reflect the views of the NSF or the NIH.
REFERENCES
[1] Aguilar, O., Prado, R., Huerta, G., and West, M. (1999). Bayesian inference
on latent structure in time series (with discussion). In Bayesian Statistics 6
(ed. J. M. Bernardo, J. O. Berger, A. P. Dawid, and A. F. M. Smith), pp. 3–26.
Oxford University Press, Oxford.
[2] Aguilar, O. and West, M. (2000). Bayesian dynamic factor models and port-
folio allocation. Journal of Business and Economic Statistics, 18, 338–357.
[3] Alspach, D. L. and Sorenson, H. W. (1972). Non-linear Bayesian estima-
tion using Gaussian sum approximations. IEEE Transactions on Automatic
Control, AC-17, 439–448.
[4] Bonassi, F. V., You, L., and West, M. (2011). Bayesian learning from marginal
data in bionetwork models. Statistical Applications in Genetics & Molecular
Biology, 10, Art 49.
[5] Cargnoni, C., Müller, P., and West, M. (1997). Bayesian forecasting of multi-
national time series through conditionally Gaussian dynamic models. Jour-
nal of the American Statistical Association, 92, 640–647.
[6] Carter, C. K. and Kohn, R. (1994). Gibbs sampling for state space models.
Biometrika, 81, 541–553.
[7] Carvalho, C. M., Johannes, M., Lopes, H. F., and Polson, N. G. (2010). Par-
ticle learning and smoothing. Statistical Science, 25, 88106.
[8] Carvalho, C. M., Lucas, J. E., Wang, Q., Chang, J., Nevins, J. R., and West,
M. (2008). High-dimensional sparse factor modelling - Applications in gene
expression genomics. Journal of the American Statistical Association, 103,
1438–1456.
[9] Carvalho, C. M. and West, M. (2007). Dynamic matrix-variate graphical
models. Bayesian Analysis, 2, 69–98.
[10] Carvalho, C. M. and West., M. (2007). Dynamic matrix-variate graphical
models - A synopsis. In Bayesian Statistics 8 (ed. J. M. Bernardo, M. J. Ba-
yarri, J. O. Berger, A. P. Dawid, D. Heckerman, A. F. M. Smith, and M. West),
pp. 585–590. Oxford University Press.
[11] Chen, R. and Liu, J. S. (2000). Mixture Kalman filters. Journal of the Royal
Statistical Society B, 62(3), 493–508.
[12] Chib, S., Omori, Y., and Asai, M. (2009). Multivariate stochastic volatility.
In Handbook of Financial Time Series (ed. T. G. Andersen, R. A. Davis, J. P.
Kreiss, and T. Mikosch), New York, pp. 365–400. Springer-Verlag.
[13] Dobra, A., Jones, B., Hans, C., Nevins, J. R., and West, M. (2004). Sparse
graphical models for exploring gene expression data. Journal of Multivariate
Analysis, 90, 196–212.
[14] Doucet, A., de Freitas, N., and Gordon, N. J. (2001). Sequential Monte
Carlo Methods in Practice. Springer-Verlag, New York.
References 25
[15] Doucet, A., Godsill, S. J., and West, M. (2000). Monte Carlo filtering and
smoothing with application to time-varying spectral estimation. Proceedings
of the IEEE International Conference on Acoustics, Speech and Signal Process-
ing, II, 701–704.
[16] Fong, W., Godsill, S. J., Doucet, A., and West, M. (2002). Monte Carlo
smoothing with application to speech enhancement. IEEE Trans. Signal Pro-
cessing, 50, 438–449.
[17] Fox, E. B. and West, M. (2011). Autoregressive models for variance matri-
ces: Stationary inverse Wishart processes. Discussion Paper 11-15, Depart-
ment of Statistical Science, Duke University. Submitted for publication.
[18] Frühwirth-Schnatter, S. (1994). Data augmentation and dynamic linear
models. Journal of Time Series Analysis, 15, 183–202.
[19] Frühwirth-Schnatter, S. (2006). Finite Mixture and Markov Switching Mod-
els. Springer-Verlag, New York.
[20] Godsill, S. J., Doucet, A., and West, M. (2004). Monte Carlo smoothing for
non-linear time series. Journal of the American Statistical Association, 99,
156–168.
[21] Gordon, N. J., Salmond, D., and Smith, A. F. M. (1993). Novel approach to
nonlinear/non-Gaussian Bayesian state estimation. Proc. IEE F, 140, 107–
113.
[22] Gourieroux, C., Jasiak, J., and Sufana, R. (2009). The Wishart autore-
gressive process of multivariate stochastic volatility. Journal of Economet-
rics, 150, 167–181.
[23] Harrison, P. J. and Stevens, C. (1976). Bayesian forecasting (with discus-
sion). Journal of the Royal Statististical Society, Series B, 38, 205–247.
[24] Huerta, G. and West, M. (1999). Bayesian inference on periodicities and
component spectral structures in time series. Journal of Time Series Analy-
sis, 20, 401–416.
[25] Jones, B., Dobra, A., Carvalho, C. M., Hans, C., Carter, C., and West, M.
(2005). Experiments in stochastic computation for high-dimensional graph-
ical models. Statistical Science, 20, 388–400.
[26] Kim, S., Shephard, N., and Chib, S. (1998). Stochastic volatility: Like-
lihood inference and comparison with ARCH models. Review of Economic
Studies, 65, 361–393.
[27] Krystal, A. D., Prado, R., and West, M. (1999). New methods of time series
analysis for non-stationary EEG data: Eigenstructure decompositions of time
varying autoregressions. Clinical Neurophysiology, 110, 1–10.
[28] Krystal, A. D., Zoldi, S., Prado, R., Greenside, H. S., and West, M. (1999).
The spatiotemporal dynamics of generalized tonic-clonic seizure EEG data:
Relevance to the climinal practice of electroconvulsive therapy. In Nonlinear
Dynamics and Brain Functioning (ed. N. Pradhan, P. Rapp, and R. Sreeni-
vasan). New York: Novascience.
[29] Lauritzen, S. L. (1996). Graphical Models. Clarendon Press, Oxford.
[30] Liu, F. and West, M. (2009). A dynamic modelling strategy for Bayesian
26 References
[62] West, M. (1986). Bayesian model monitoring. Journal of the Royal Statis-
tical Society (Ser. B), 48, 70–78.
[63] West, M. (1992). Modelling with mixtures (with discussion). In Bayesian
Statistics 4 (ed. J. M. Bernardo, J. O. Berger, A. P. Dawid, and A. F. M. Smith),
pp. 503–524. Oxford University Press.
[64] West, M. (1993). Approximating posterior distributions by mixtures. Jour-
nal of the Royal Statistical Society (Ser. B), 54, 553–568.
[65] West, M. (1993). Mixture models, Monte Carlo, Bayesian updating and
dynamic models. Computing Science and Statistics, 24, 325–333.
[66] West, M. (1995). Bayesian inference in cyclical component dynamic linear
models. Journal of the American Statistical Association, 90, 1301–1312.
[67] West, M. (1996). Bayesian time series: Models and computations for the
analysisd of time series in the physical sciences. In Maximum Entropy and
Bayesian Methods 15 (ed. K. Hanson and R. Silver), pp. 23–34. Kluwer.
[68] West, M. (1996). Some statistical issues in Palæoclimatology (with discus-
sion). In Bayesian Statistics 5 (ed. J. M. Bernardo, J. O. Berger, A. P. Dawid,
and A. F. M. Smith), pp. 461–486. Oxford University Press.
[69] West, M. (1997). Modelling and robustness issues in Bayesian time series
analysis (with discussion). In Bayesian Robustness (ed. J. O. Berger, B. Betrò,
E. Moreno, L. R. Pericchi, F. Ruggeri, G. Salinetti, and L. Wasserman), IMS
Monographs, pp. 231–252. Institute for Mathematical Statistics.
[70] West, M. (1997). Time series decomposition. Biometrika, 84, 489–494.
[71] West, M. (2003). Bayesian factor regression models in the “large p, small
n” paradigm. In Bayesian Statistics 7 (ed. J. M. Bernardo, M. J. Bayarri,
J. O. Berger, A. P. David, D. Heckerman, A. F. M. Smith, and M. West), pp.
723–732. Oxford University Press.
[72] West, M. and Harrison, P. J. (1986). Monitoring and adaptation in Bayesian
forecasting models. Journal of the American Statistical Association, 81, 741–
750.
[73] West, M. and Harrison, P. J. (1989). Subjective intervention in formal
models. Journal of Forecasting, 8, 33–53.
[74] West, M. and Harrison, P. J. (1997). Bayesian Forecasting & Dynamic Models
(2nd edn). Springer Verlag.
[75] West, M., Prado, R., and Krystal, A. D. (1999). Evaluation and comparison
of EEG traces: Latent structure in non-stationary time series. Journal of the
American Statistical Association, 94, 1083–1095.
[76] Wilkinson, D. J. (2006). Stochastic Modelling for Systems Biology. London:
Chapman & Hall/CRC.
[77] Wilkinson, D. J. (2011). Parameter inference for stochastic kinetic models
of bacterial gene regulation: a Bayesian approach to systems biology (with
discussion). In Bayesian Statistics 9 (ed. J. M. Bernardo, M. J. Bayarri, J. O.
Berger, A. P. David, D. Heckerman, A. F. M. Smith, and M. West), pp. 679–
700. Oxford University Press.
[78] Yao, G., Tan, C., West, M., Nevins, J. R., and You, L. (2011). Origin of bista-
References 29