Abstract
Linear stochastic differential equations are expressed as an exact discrete model (EDM) and estimated with structural equation models (SEMs) and the Kalman filter (KF) algorithm. The oversampling approach is introduced in order to formulate the EDM on a time grid which is finer than the sampling intervals. This leads to a simple computation of the nonlinear parameter functionals of the EDM. For small discretization intervals, the functionals can be linearized, and standard software permitting only linear parameter restrictions can be used. However, in this case the SEM approach must handle large matrices leading to degraded performance and possible numerical problems. The methods are compared using coupled linear random oscillators with time-varying parameters and irregular sampling times.
Acknowledgments
I would like to thank the referees for detailed comments which helped me to improve the presentation of the article.
Notes
1The intercepts
are obtained from the interpolated exogenous variables
. A more elaborated nonlinear treatment of (14–16) is given in Appendix B.
2The model has numerous applications, for example, the statistical analysis of sunspots (Bartlett, 1946; Arato, Kolmogoroff, & Sinai, Citation1962; H. Singer, Citation1993), in physics (for an overview see Gitterman, Citation2005) or in psychology (coupled oscillators; Boker & Poponak, Citation2004).
3The formal derivative ζ(t) = dW(t)/dt is the white noise process (cf. Arnold, Citation1974).
Note. Top: Exact EDM (Δ
t
= .5). Bottom: Linearized EDM, oversampling (J = 5; δt = .1). SEM = structural equation model; KF = Kalman filter; ML = maximum likelihood; EDM = exact discrete model.
6One may interpret the system as interaction of married couples over time.
Note. ML = maximum likelihood; SEM = structural equation model; KF = Kalman filter.
Note. Oversampling times δt = 0.1, 0.05, 0.025. Numbers rounded to four digits. The likelihoods and evaluation times are shown in the last two rows. ML = maximum likelihood; SEM = structural equation model; KF = Kalman filter.
Note. Oversampling times δt = 0.1, 0.05, 0.025. Numbers rounded to four digits. The likelihoods and evaluation times are shown in the last two rows. ML = maximum likelihood; SEM = structural equation model; KF = Kalman filter.
7cf. e.g., Langevin (Citation1908), Uhlenbeck and Ornstein (Citation1954), Bartlett (Citation1946, 1955/1978), Itô (Citation1951), Stratonovich (Citation1960), Kalman and Bucy (Citation1961), Schweppe (Citation1965), Nelson (Citation1967), Bergstrom (Citation1976a), Coleman (Citation1968), Jazwinski (Citation1970), Black and Scholes (Citation1973), Jennrich and Bright (Citation1976), Phillips (Citation1976a), Doreian and Hummon (Citation1976, Citation1979), Jones (Citation1984), Moebus and Nagl (1983), Arminger (Citation1986), H. Singer (Citation1986, Citation1990, Citation1992b, 1998b), Zadrozny (Citation1988), Hamerle et al. (Citation1991), Hamerle et al. (Citation1993), Oud and Jansen (Citation2000), Oud and Singer (Citation2008a, 2008b).
Arato
,
M.
,
Kolmogoroff
,
A.
, &
Sinai
,
J.
(
1962
).
Estimation of the parameters of a complex stationary Gaussian Markov process
.
Soviet Mathematics Doklady
,
3
,
1368
–
1371
.
Singer
,
H.
(
1993
).
Continuous-time dynamical systems with sampled data, errors of measurement and unobserved components
.
Journal of Time Series Analysis
,
14
,
527
–
545
.
Gitterman
,
M.
(
2005
).
The noisy oscillator
.
Singapore
:
World Scientific
.
Boker
,
S. M.
, &
Poponak
,
S.
(
2004
).
Multivariate multilevel models of dynamical systems using latent differential equations
.
In
C.
van Dijkum
,
J.
Blasius
, &
C.
Durand
(Eds.)
,
Proceedings of the sixth International Conference on Social Science Methodology, Amsterdam 2004
[CD-ROM]
.
Farmington Hills, MI
:
Budrich Publishers
.
Arnold
,
L.
(
1974
).
Stochastic differential equations
.
New York, NY
:
John Wiley
.
Singer
,
B.
, &
Spilerman
,
S.
(
1976
).
Representation of social processes by Markov models
.
American Journal of Sociology
,
82
,
1
–
53
.
Phillips
,
P.
(
1976a
).
The estimation of linear stochastic differential equations with exogenous variables
. In
A.
Bergstrom
(Ed.),
Statistical inference in continuous time models
(pp.
135
–
173
).
Amsterdam, The Netherlands
:
North Holland
.
Hansen
,
L.
, &
Sargent
,
T.
(
1983
).
The dimensionality of the aliasing problem in models with rational spectral density
.
Econometrica
,
51
,
377
–
387
.
Hamerle
,
A.
,
Nagl
,
W.
, &
Singer
,
H.
(
1991
).
Problems with the estimation of stochastic differential equations using structural equations models
.
The Journal of Mathematical Sociology
,
16
,
201
–
220
.
Hamerle
,
A.
,
Nagl
,
W.
, &
Singer
,
H.
(
1993
).
Identification and estimation of continuous time dynamic systems with exogenous variables using panel data
.
Econometric Theory
,
9
,
283
–
295
.
Singer
,
H.
(
1992a
).
The aliasing phenomenon in visual terms
.
The Journal of Mathematical Sociology
,
17
,
39
–
49
.
Langevin
,
P.
(
1908
).
Sur la théorie du mouvement Brownien [On the theory of Brownian motion]
.
Comptes Rendus de l'Academie des Sciences (Paris)
,
146
,
530
–
533
.
Uhlenbeck
,
G.
, &
Ornstein
,
L.
(
1954
).
On the theory of Brownian motion
. In
N.
Wax
(Ed.),
Selected papers on noise and stochastic processes
(pp.
93
–
111
).
Mineola, NY
:
Dover
.
Bartlett
,
M.
(
1946
).
On the theoretical specification and sampling properties of autocorrelated time-series
.
Journal of the Royal Statistical Society (Supplement)
,
7
,
27
–
41
.
Itô
,
K.
(
1951
).
On stochastic differential equations
.
Memoirs of the American Mathematical Society
,
4
,
1
–
51
.
Stratonovich
,
R.
(1960). Conditional Markov processes.
Theory of Probability and Its Applications
, 5, 156–178.
Kalman
,
R.
, &
Bucy
,
R.
(
1961
).
New results in linear filtering and prediction theory
.
Transactions of the ASME, Series D: Journal of Basic Engineering
,
83
,
95
–
108
.
Schweppe
,
F.
(
1965
).
Evaluation of likelihood functions for gaussian signals
.
IEEE Transactions on Information Theory
,
11
,
61
–
70
.
Nelson
,
E.
(
1967
).
Dynamical theories of Brownian motion
.
Princeton, NJ
:
Princeton University
.
Bergstrom
,
A.
(
1976a
).
Non recursive models as discrete approximations to systems of stochastic differential equations
. In
A.
Bergstrom
(Ed.),
Statistical inference in continuous time models
(pp.
15
–
26
).
Amsterdam, The Netherlands
:
North Holland
.
Coleman
,
J.
(
1968
).
The mathematical study of change
. In
H.
Blalock
&
A. B.
Blalock
(Eds.),
Methodology in social research
(pp.
428
–
478
).
New York, NY
:
McGraw-Hill
.
Jazwinski
,
A.
(
1970
).
Stochastic processes and filtering theory
.
New York, NY
:
Academic Press
.
Black
,
F.
, &
Scholes
,
M.
(
1973
).
The pricing of options and corporate liabilities
.
Journal of Political Economy
,
81
,
637
–
654
.
Jennrich
,
R.
, &
Bright
,
P.
(
1976
).
Fitting systems of linear differential equations using computer generated exact derivatives
.
Technometrics
,
18
,
385
–
392
.
Phillips
,
P.
(
1976a
).
The estimation of linear stochastic differential equations with exogenous variables
. In
A.
Bergstrom
(Ed.),
Statistical inference in continuous time models
(pp.
135
–
173
).
Amsterdam, The Netherlands
:
North Holland
.
Doreian
,
P.
, &
Hummon
,
N.
(
1976
).
Modelling social processes
.
New York, NY
:
Elsevier
.
Doreian
,
P.
, &
Hummon
,
N.
(
1979
).
Estimating differential equation models on time series: Some simulation evidence
.
Sociological Methods and Research
,
8
,
3
–
33
.
Jones
,
R.
(
1984
).
Fitting multivariate models to unequally spaced data
. In
E.
Parzen
(Ed.),
Time series analysis of irregularly observed data
(pp.
158
–
188
).
New York, NY
:
Springer
.
Arminger
,
G.
(
1986
).
Linear stochastic differential equation models for panel data with unobserved variables
.
In
N.
Tuma
(Ed.)
,
Sociological methodology
(Vol.
16
, pp.
187
–
212
).
Washington, DC
:
American Sociological Association
.
Singer
,
H.
(
1986
).
Depressivität und gelernte Hilflosigkeit als Stochastischer Prozeß [Depression and learned helplessness as stochastic process] (Unpublished master's thesis)
.
Universität Konstanz
,
Konstanz
,
Germany
.
Singer
,
H.
(
1990
).
Parameterschätzung in zeitkontinuierlichen dynamischen Systemen [Parameter estimation in continuous time dynamical systems] (Doctoral dissertation)
.
Konstanz
,
Germany
:
Hartung-Gorre-Verlag
.
Singer
,
H.
(
1992b
).
Zeitkontinuierliche Dynamische Systeme [Continuous time dynamical systems]
.
Frankfurt a.M.
,
Germarny
:
Campus-Verlag
.
Bergstrom
,
A.
(Ed.). (
1976b
).
Statistical inference in continuous time models
.
Amsterdam, The Netherland
:
North Holland
.
Hamerle
,
A.
,
Nagl
,
W.
, &
Singer
,
H.
(
1991
).
Problems with the estimation of stochastic differential equations using structural equations models
.
The Journal of Mathematical Sociology
,
16
,
201
–
220
.
Hamerle
,
A.
,
Nagl
,
W.
, &
Singer
,
H.
(
1993
).
Identification and estimation of continuous time dynamic systems with exogenous variables using panel data
.
Econometric Theory
,
9
,
283
–
295
.
Oud
,
J.
, &
Jansen
,
R.
(
2000
).
Continuous time state space modeling of panel data by means of SEM
.
Psychometrika
,
65
,
199
–
215
.
Oud
,
J.
, &
Singer
,
H.
(
2008a
).
Continuous time modeling of panel data: SEM vs. filter techniques
.
Statistica Neerlandica
,
62
(
1
),
4
–
28
.