# Spectral density-based and measure-preserving ABC for partially observed diffusion processes. An illustration on Hamiltonian SDEs

## Abstract

Approximate Bayesian computation (ABC) has become one of the major tools of likelihood-free statistical inference in complex mathematical models. Simultaneously, stochastic differential equations (SDEs) have developed to an established tool for modelling time-dependent, real-world phenomena with underlying random effects. When applying ABC to stochastic models, two major difficulties arise: First, the derivation of effective summary statistics and proper distances is particularly challenging, since simulations from the stochastic process under the same parameter configuration result in different trajectories. Second, exact simulation schemes to generate trajectories from the stochastic model are rarely available, requiring the derivation of suitable numerical methods for the synthetic data generation. To obtain summaries that are less sensitive to the intrinsic stochasticity of the model, we propose to build up the statistical method (e.g. the choice of the summary statistics) on the underlying structural properties of the model. Here, we focus on the existence of an invariant measure and we map the data to their estimated invariant density and invariant spectral density. Then, to ensure that these model properties are kept in the synthetic data generation, we adopt measure-preserving numerical splitting schemes. The derived property-based and measure-preserving ABC method is illustrated on the broad class of partially observed Hamiltonian type SDEs, both with simulated data and with real electroencephalography data. The derived summaries are particularly robust to the model simulation, and this fact, combined with the proposed reliable numerical scheme, yields accurate ABC inference. In contrast, the inference returned using standard numerical methods (Euler–Maruyama discretisation) fails. The proposed ingredients can be incorporated into any type of ABC algorithm and directly applied to all SDEs that are characterised by an invariant distribution and for which a measure-preserving numerical method can be derived.

## Keywords

Approximate Bayesian computation Likelihood-free inference Stochastic differential equations Numerical splitting schemes Invariant measure Neural mass models## 1 Introduction

Over the last decades, stochastic differential equations (SDEs) have become an established and powerful tool for modelling time-dependent, real-world phenomena with underlying random effects. They have been successfully applied to a variety of scientific fields, ranging from biology over finance to physics, chemistry, neuroscience and others. Diffusion processes obtained as solutions of SDEs are typically characterised by some underlying structural properties whose investigation and preservation are crucial. Examples are boundary properties, symmetries or the preservation of invariants or qualitative behaviour such as the ergodicity or the conservation of energy. Here, we focus on a specific structural property, namely the existence of a unique invariant measure. Besides the modelling, it is of primary interest to estimate the underlying model parameters. This is particularly difficult when the multivariate stochastic process is only partially observed through a 1-dimensional function of its coordinates (the output process), a scenario that we tackle here. Moreover, due to the complexity of SDEs, needed to understand and reproduce the real data, the underlying likelihood is often unknown or intractable. Among several likelihood-free inference approaches, we focus on the simulation-based approximate Bayesian computation (ABC) method. We refer to Marin et al. (2012) and to the recently published book “Handbook of approximate Bayesian computation” for an exhaustive discussion (Sisson et al. 2018).

ABC has become one of the major tools for parameter inference in complex mathematical models in the last decade. The method is based on the idea of deriving an approximate posterior density targeting the true (unavailable) posterior by running massive simulations from the model to replace the intractable likelihood. It was first introduced in the context of population genetics; see, e.g. Beaumont et al. (2002). Since then, it has been successfully applied in a wide range of fields; see, e.g. Barnes et al. (2012), Blum (2010a), Boys et al. (2008), McKinley et al. (2017), Moores et al. (2015) and Toni et al. (2009). Moreover, ABC has also been proposed to infer parameters from time series models (see, e.g. Drovandi et al. 2016; Jasra 2015), state space models (see, e.g. Martin et al. 2019; Tancredi 2019) and SDE models (see, e.g. Kypraios et al. 2017; Maybank et al. 2017; Picchini 2014; Picchini and Forman 2016; Picchini and Samson 2018; Sun et al. 2015; Zhu et al. 2016). Several advanced ABC algorithms have been proposed in the literature, such as sequential Monte Carlo (SMC) ABC, Markov Chain Monte Carlo (MCMC) ABC, sequential annealing ABC, noisy ABC; see, e.g. Fan and Sisson (2018) and the references therein for a recent review. The idea of the basic acceptance–rejection ABC algorithm is to keep a sampled parameter value from the prior as a realisation from the approximate posterior, if the distance between the summary statistics of the synthetic dataset, which is generated conditioned on this parameter value, and the summaries of the original reference data is smaller than some tolerance level. The goal of this paper is to illustrate how building up the ABC method on the structural properties of the underlying SDE, and using a numerical method capable of preserving them in the generation of the data from the model, leads to a successful inference even when applying ABC in this basic acceptance–rejection form.

The performance of any ABC method depends heavily on the choice of “informative enough” summary statistics, a suitable distance measure and a tolerance level \(\epsilon \). The quality of the approximation improves as \(\epsilon \) decreases, and it has been shown that, under some conditions, the approximated ABC posterior converges to the true one when \(\epsilon \rightarrow 0\) (Jasra 2015). At the same time though, the computational cost increases when \(\epsilon \) decreases. A possibility is to use ad hoc threshold selection procedures; see, e.g. Barber et al. (2015), Blum (2010b), Lintusaari et al. (2017), Prangle et al. (2014) and Robert (2016). Here, we fix the tolerance level \(\epsilon \) as a percentile of the calculated distances. This is another common practice known as “reference table acceptance–rejection ABC” (Cornuet et al. 2008) and used, for example, in Beaumont et al. (2002), Biau et al. (2015), Sun et al. (2015) and Vo et al. (2015). Instructions for constructing effective summaries and distances are rare, and they depend on the problem under consideration; see, e.g. Fearnhead and Prangle (2012) for a semiautomatic linear regression approach, Jiang et al. (2017) for an automatic construction approach based on training deep neural networks and Blum (2010b) and Prangle (2018) for two recent reviews. To avoid the information loss caused by using non-sufficient summary statistics, another common procedure is to work with the entire dataset; see, e.g. Jasra (2015) and Sun et al. (2015). This requires the application of more sophisticated distances *d* such as the Wasserstein metric (Bernton et al. 2019; Muskulus and Verduyn-Lunel 2011) or other distances designed for time series; for an overview see, e.g. Mori et al. (2016) and the references therein.

When working with stochastic models, simulations from the stochastic simulator, conditionally to the same parameter configuration, yield different trajectories. To consider summary statistics that are less sensitive to the intrinsic stochasticity of the model (Wood 2010), we choose them based on the structural property of an underlying invariant measure. The idea is to map the data, i.e. the realisations of the output process, to an object that is invariant for repeated simulations under the same parameter setting and is instead sensitive to small changes in the parameters. In particular, we map the data to their estimated invariant density and invariant spectral density, taking thus the dependence structure of the dynamical model into account. The distance measure can then be chosen according to the mapped data.

As other simulation-based statistical methods, e.g. MCMC, SMC or machine learning algorithms, ABC relies on the ability of simulating data from the model. However, the exact simulation from complex stochastic models is rarely possible, and thus, numerical methods need to be applied. This introduces a new level of approximation into the ABC framework. When the statistical method is build upon the structural properties of the underlying model, the successful inference can only be guaranteed when these properties are preserved in the synthetic data generated from the model. However, the issue of deriving a property-preserving numerical method when applying ABC to SDEs is usually seen as not so relevant, and it is usually recommended to use the Euler–Maruyama scheme or one of the higher-order approximation methods described in Kloeden and Platen (1992); see, e.g. Picchini (2014), Picchini and Forman (2016), Picchini and Samson (2018) and Sun et al. (2015). In general, these standard methods do not preserve the underlying structural local and global properties of the model; see, e.g. Ableidinger et al. (2017), Malham and Wiese (2013), Moro and Schurz (2007) and Strømmen Melbø and Higham (2004).

Here, we propose to apply structure-preserving numerical splitting schemes within the ABC algorithm. The idea of these methods is to split the SDE into explicitly solvable subequations and to apply a proper composition of the resulting exact solutions. Standard procedures are, for example, the Lie–Trotter method and the usually more accurate Strang approach; see, e.g. Leimkuhler et al. (2016). Since the only approximation enters through the composition of the derived explicit solutions, numerical splitting schemes usually preserve the structural properties of the underlying SDE and accurately reproduce its qualitative behaviour. Moreover, they usually have the same order of convergence as the frequently applied Euler–Maruyama method and are likewise efficient. We refer to Blanes et al. (2009) and Mclachlan and Quispel (2002) for an exhaustive discussion of splitting methods for broad classes of ordinary differential equations (ODEs), which partially have already been carried over to SDEs; see, e.g. Misawa (2001) for a general class of SDEs, Ableidinger and Buckwar (2016) for the stochastic Landau–Lifshitz equations, Bréhier and Goudenège (2019) for the Allen–Cahn equation and Ableidinger et al. (2017) for Hamiltonian type SDEs.

The main contribution of this work lies in the combination of the proposed invariant measure-based summary statistics and the measure-preserving numerical splitting schemes within the ABC framework. We demonstrate that a simulation-based inference method, here ABC, can only perform well if the underlying simulation method preserves the structural properties of the SDE. While the use of preserving splitting schemes within the ABC method yields successful results, applying a general purpose numerical method, such as the Euler–Maruyama discretisation, may result in seriously wrong inferences. We illustrate the proposed spectral density-based and measure-preserving ABC method on the class of stochastic Hamiltonian type equations for which the existence of an underlying unique invariant distribution and measure-preserving numerical splitting schemes have been already intensively studied in the literature; see, e.g. Ableidinger et al. (2017), Mattingly et al. (2002), Leimkuhler and Matthews (2015) and Milstein and Tretyakov (2004). Hamiltonian type SDEs have been investigated in molecular dynamics, where they are typically referred to as Langevin equations; see, e.g. Leimkuhler and Matthews (2015). Recently, they have also received considerable attention in the field of neuroscience as the so-called neural mass models (Ableidinger et al. 2017).

The paper is organised as follows: In Sect. 2, we recall the acceptance–rejection ABC setting. We introduce the invariant measure-based summary statistics and propose a proper distance. We then discuss the importance of considering measure-preserving numerical schemes for the synthetic data generation when exact simulation methods are not applicable and provide a short introduction to numerical splitting methods. In Sect. 3, we introduce Hamiltonian type SDEs and recall two splitting integrators preserving the invariant measure of the model. In Sect. 4, we validate the proposed method by investigating the stochastic harmonic oscillator, for which exact simulation is possible. In Sect. 5, we apply the proposed ABC method to the stochastic Jansen and Rit neural mass model (JR-NMM). We refer to Jansen and Rit (1995) for the original version, an ODE with a stochastic input function, and to Ableidinger et al. (2017) for its reformulation as a Hamiltonian type SDE. This model has been reported to successfully reproduce electroencephalography (EEG) data. We illustrate the performance of the proposed ABC method with both simulated and real data. Final remarks, possible extensions and conclusions are reported in Sect. 6. A supplementary material (Online Resource 1) with further illustrations of the proposed method is available online, and a sample code used to generate the main results is available on github at https://github.com/massimilianotamborrino/sdbmpABC.

## 2 Spectral density-based and measure-preserving ABC for partially observed SDEs with an invariant distribution

*n*-dimensional, \(n \in {\mathbb {N}}\), non-autonomous SDE of Itô-type describing the time evolution of a system of interest

*r*-dimensional, \(r \in {\mathbb {N}}\), Wiener process with independent and \({\mathbb {F}}\)-adapted components. We further assume that the drift component \(f:~[0,T] \times {\mathbb {R}}^n \rightarrow {\mathbb {R}}^n\) and the diffusion component \({\mathcal {G}}: [0,T] \times {\mathbb {R}}^n \rightarrow {\mathbb {R}}^{n \times r}\) fulfil the necessary global Lipschitz and linear growth conditions, such that the existence and the pathwise uniqueness of an \({\mathbb {F}}\)-adapted strong solution process \(\mathbf{X }=(X(t))_{t \in [0,T]} \in {\mathbb {R}}^n\) of (1) are guaranteed; see, e.g. Arnold (1974).

*n*-dimensional solution process \(\mathbf{X }\) is only partially observed through the 1-dimensional and parameter-dependent output process

**X**is partially observed through discrete time measurements of the output process \(\mathbf{Y }_\theta \) given in (2), by benefiting from the (in general unknown) invariant distribution \(\eta _{\mathbf{Y }_\theta }\) satisfying (3).

### 2.1 The ABC method

*s*is a vector of sufficient statistics for \(\theta \), the acceptance–rejection ABC produces samples from the true posterior \(\pi (\theta |y)\). Due to the complexity of the underlying SDE (1), we cannot derive non-trivial sufficient statistics

*s*for \(\theta \). Moreover, due to the underlying stochasticity of the model, \({\mathbb {P}}(d(s(y),s(y_{\theta '}))=0)=0\). Thus, \(\epsilon \) is required to be strictly positive.

*p*of the calculated distances \(D_i\). Algorithm 1 yields samples from an approximated posterior \(\pi _{\text {ABC}}(\theta |y)\) according to

Besides the tolerance level \(\epsilon \), the quality of the ABC method depends strongly on the choice of suitable summary statistics combined with a proper distance measure and on the numerical method used to generate the synthetic data from the model. In the following, we introduce summaries that are very effective for the class of models having an underlying invariant distribution, we suggest a proper distance based on them, and we propose the use of measure-preserving numerical splitting schemes.

### 2.2 An effective choice of summaries and distances: spectral density-based ABC

When applying ABC to stochastic models, an important statistical challenge arises. Due to the intrinsic randomness, repeated simulations of the process \(\mathbf{Y }_\theta \) under the same parameter vector \(\theta \) may yield very different trajectories. An illustration is given in Fig. 1 (top and middle panels), where we report two trajectories of the output process of the stochastic JR-NMM (25) generated with an identical parameter configuration. This model is a specific SDE of type (1), observed through \(\mathbf{Y }_\theta \) as in (2), and admitting an invariant distribution \(\eta _{\mathbf{Y }_{\theta }}\) satisfying (3). See Sect. 5 for a description of the model. In the top panel, we visualise the full paths for a time \(T=200\), while in the middle panel we provide a zoom, showing only the initial part.

### Proposal 1

To use the property of an invariant measure \(\eta _{{{\mathbf{Y}}}_\theta }\) and to map the data \(y_\theta \) to their estimated invariant density \({\hat{f}}_{y_\theta }\) and invariant spectral density \({\hat{S}}_{y_\theta }\).

*s*of a dataset \(y_\theta \) as

*f*-divergences (see, e.g. Sason and Verdú 2016), or the Wasserstein distance, recently proposed for ABC (Bernton et al. 2019). Within the ABC framework (see Step 7 in Algorithm 1), we suggest to use the following distance

*L*times:

Then, we take the median of the resulting *L* values \(w'\). See, e.g. Prangle (2017) for alternative approaches for the derivation of weights among summary statistics. Since the densities \({{\hat{f}}}_{{y_\theta }}\) and \({{\hat{S}}}_{y_\theta }\) are estimated at discrete points, the IAE (6) is approximated applying trapezoidal integration.

*M*datasets

*y*as a long-time trajectory (when using simulated observed reference data) or as a long-time recording of the modelled phenomenon (when using real observed reference data) that is cut into

*M*pieces. Alternatively,

*y*would consist of M independent repeated experiments or simulations, when dealing with real or simulated data, respectively. As expected, having \(M>1\) datasets improves the quality of the estimation due to the increased number of observations.

### 2.3 A new proposal of synthetic data generation: measure-preserving ABC

### Proposal 2

To adopt a numerical method for the synthetic data generation that preserves the underlying invariant measure of the model.

*f*and the diffusion \({\mathcal {G}}\) of SDE (1) can be written as

*f*and \({\mathcal {G}}\) in a way such that the resulting

*d*subequations

*t*and starting from \(X_0\). Once these explicit solutions are derived, a proper composition needs to be applied. Here we use the Strang approach (Mclachlan and Quispel 2002; Strang 1968)

In Fig. 2, we illustrate how the numerical splitting method preserves the underlying invariant measure of the weakly damped stochastic harmonic oscillator (23), independently from the choice of the time step \(\varDelta \). This is a specific SDE of type (1), observed through \(\mathbf{Y }_\theta \) as in (2) and with a known invariant distribution \(\eta _{\mathbf{Y }_{\theta }}\). See Sect. 3 for the detailed numerical splitting scheme and Sect. 4 for a description of the model. In contrast, the Euler–Maruyama scheme performs worse as \(\varDelta \) increases. Each subplot shows a comparison of the true invariant density (blue solid lines) and the corresponding kernel estimate \({\hat{f}}_{y_\theta }\) based on a path \(y_\theta \) from the model, generated from the measure-preserving numerical splitting scheme (22) (orange dashed lines) or the Euler–Maruyama approach (green dotted lines). The data are generated under \(T=10^3\) and different values for the time step, namely \(\varDelta =10^{-3}\), \(3\cdot 10^{-3}\), \(4.5\cdot 10^{-3}\).

### 2.4 Notation

We apply the summary statistics (5) and the distance (8) in Algorithm 1. We use the notation Algorithm 1 (i) [spectral density-based ABC method] when the synthetic data are simulated exactly, Algorithm 1 (ii) [spectral density-based and measure-preserving ABC method] when a measure-preserving numerical splitting scheme is applied and Algorithm 1 (iii) when we generate the data with the non-preserving Euler–Maruyama scheme.

To evaluate the performance of the proposed ABC method, we analyse the marginal posterior densities, denoted by \(\pi _{\text {ABC}}^*(\theta _j|y)\), \(j \in \{1,\ldots ,k\}\), obtained from the posterior density \(\pi _{\text {ABC}}^*(\theta |y)\) corresponding to \(\pi _{\text {ABC}}(\theta |y)\), \(\pi _{\text {ABC}}^\text {num}(\theta |y)\) or \(\pi _{\text {ABC}}^e(\theta |y)\), depending on whether we obtain it from Algorithm 1 (i), (ii) or (iii). Following this notation, we define by \({\hat{\theta }}_{\text {ABC},j}^*\) the marginal ABC posterior means.

## 3 An illustration on Hamiltonian type SDEs

*n*-dimensional (\(n=2d\), \(d\in {\mathbb {N}}\)) stochastic process

*d*-dimensional components

*n*-dimensional SDE of Hamiltonian type with initial value \(X_0=~(Q_0,P_0)^{\prime }\) and

*d*-dimensional (\(r=d\)) Wiener process \(\mathbf{W }\) describes the time evolution of \(\mathbf{X }\) by

*Q*and

*P*, respectively. The SDE (10) consists of 4 parts, each representing a specific type of behaviour. In this configuration, the first is the

*Hamiltonian part*involving \(H:{\mathbb {R}}^d \times {\mathbb {R}}^d \rightarrow {\mathbb {R}}_0^+\) given by

*linear damping part*, described by the diagonal matrix \(\varGamma _\theta =\text {diag}[\gamma _{1},\ldots ,\gamma _{d}] \in {\mathbb {R}}^{d \times d}\). The third is the

*nonlinear displacement part*, consisting of the nonlinear and globally Lipschitz continuous function \(G: {\mathbb {R}}^d \rightarrow {\mathbb {R}}^d\). The fourth corresponds to the

*diffusion part*, given by \(\varSigma _\theta =\text {diag}[\sigma _{1},\ldots ,\sigma _{d}] \in {\mathbb {R}}^{d \times d}\).

### 3.1 Structural model property

Under the requirement of non-degenerate matrices \(\varLambda _\theta \), \(\varGamma _\theta \) and \(\varSigma _\theta \), i.e. strictly positive diagonal entries, Hamiltonian type SDEs as in (10) are often ergodic. As a consequence, the distribution of the solution process \(\mathbf{X }\) (and thus of the output process \(\mathbf{Y}_\theta \)) converges exponentially fast towards a unique invariant measure \(\eta _{\mathbf{X }}\) on \(({\mathbb {R}}^n,{\mathcal {B}}({\mathbb {R}}^n))\) (and thus \(\eta _{{{\mathbf{Y}}}_\theta }\) on \(({\mathbb {R}},{\mathcal {B}}({\mathbb {R}}))\); see, e.g. Ableidinger et al. (2017) and the references therein.

### 3.2 Measure-preserving numerical splitting schemes

*G*, the SDE (10) cannot be solved explicitly. With the purpose of excluding

*G*, the Hamiltonian type SDE (10) is split into the two subsystems

*d*-dimensional zero vector. This results in a linear SDE with additive noise (11) and a nonlinear ODE (12) that can be both explicitly solved. Indeed, since \(\nabla _P H(Q(t),P(t))=P(t)\) and \(\nabla _Q H(Q(t),P(t))\!=\!\varLambda _\theta ^2 Q(t)\), Subsystem (11) can be rewritten as

*n*-dimensional Gaussian vectors with null mean and variance \(C(\varDelta )\), where the matrix

*C*(

*t*) follows the dynamics of the matrix-valued ODE

*G*depends only on the component \(\mathbf{Q }\), the exact path of Subsystem (12) is obtained through

*d*-dimensional Gaussian vectors with null mean and variance \(\varDelta {\mathbb {I}}_d\). The Strang approach is now given by

### 3.3 Implementation details

The ABC procedure is coded in the computing environment **R** (R Development Core Team 2011), using the package **Rcpp** (Eddelbuettel and François 2011), which offers a seamless integration of **R** and C++, drastically reducing the computational time of the algorithms. The code is then parallelised using the R-packages foreach and doParallel, taking advantage of the *for loop* in the algorithm. All simulations are run on the HPC cluster RADON1, a high-performing multiple core cluster located at the Johannes Kepler University Linz. To obtain smoothed periodogram estimates, we apply the R-function spectrum. It requires the specification of a smoothing parameter span. In all our experiments, we use span \(=5T\). In addition, we avoid using a logarithmic scale by setting the log parameter to “no”. To obtain kernel estimates of the invariant density, we apply the R-function density. Here, we use the default value for the smoothing bandwidth bw and set the number of points at which the invariant density has to be estimated to n\(=10^3\). The invariant spectral density is estimated at the default values of the spectrum function. A sample code is publicly available on github at https://github.com/massimilianotamborrino/sdbmpABC.

## 4 Validation of the proposed ABC method when exact simulation is possible

### 4.1 Weakly damped stochastic harmonic oscillator: the model and its properties

*A*and

*C*introduced in (13) and (15), respectively. The invariant distribution \(\eta _\mathbf{X }\) of the solution process \(\mathbf{X }\) is given by

### 4.2 Validation of the spectral density-based ABC Algorithm 1 (i)

*T*] and with the same time step as the observed data. We choose independent uniform priors, in particular,

### 4.3 Validation of the spectral density-based and measure-preserving ABC Algorithm 1 (ii)

*T*,

*w*,

*M*and

*N*as before, for different values of the time step \(\varDelta \). In particular, we choose \(\varDelta = 5\cdot 10^{-3}\) (top panels), \(\varDelta =7.5 \cdot 10^{-3}\) (middle panels) and \(\varDelta =10^{-2}\) (lower panels). The posteriors obtained from Algorithm 1 (ii) successfully targets \(\pi _{\text {ABC}}(\theta |y)\), even for a time step as large as \(\varDelta =10^{-2}\). On the contrary, Algorithm 1 (iii) is not even applicable. Indeed, the numerical scheme computationally pushes the amplitude of the oscillator towards infinity, resulting in a computer overflow, i.e. \({\widetilde{Y}}_\theta (t_i) \approx \infty \). Thus, neither \({\hat{f}}_{{{\tilde{y}}}_\theta }\) nor \({\hat{S}}_{{{\tilde{y}}}_\theta }\) can be computed and the density \(\pi _{\text {ABC}}^e(\theta |y)\) cannot be derived.

## 5 Validation of the spectral density-based and measure-preserving ABC Algorithm 1 (ii) on simulated and real data

We now illustrate the performance of Algorithm 1 (ii) by applying it to the stochastic JR-NMM. We rely on the efficient numerical splitting scheme (17) to guarantee measure-preserving synthetic data generation within the ABC framework. After estimating the parameters from simulated data, we infer them from real EEG data. In the available supplementary material, we illustrate the performance of Algorithm 1 (ii) also on the nonlinear damped stochastic oscillator, an extended version of the weakly damped harmonic oscillator discussed in Sect. 4.

### 5.1 The stochastic Jansen and Rit neural mass model

*G*are \(\mu \),

*A*,

*B*and \(C_i\), \(i=1,2,3,4\)\(\in {\mathbb {R}}^+\). The coefficients

*A*and

*B*describe the average excitatory and inhibitory synaptic gain, respectively. The parameters \(C_i\) are internal connectivity constants, which reduce to only one parameter

*C*, by using the relations \(C_1=C\), \(C_2=0.8C\), \(C_3=0.25C\) and \(C_4=0.25C\); see Jansen and Rit (1995).

### 5.2 Parameter inference from simulated data

Not all model parameters of the JR-NMM are of biological interest or can be simultaneously identified. For example, the noise coefficients \(\sigma _4\) and \(\sigma _6\) were introduced mainly for mathematical convenience in Ableidinger et al. (2017). To guarantee the existence of a unique invariant measure \(\eta _{\mathbf{X }}\) on \(({\mathbb {R}}^6,{\mathcal {B}}({\mathbb {R}}^6))\), they are required to be strictly positive. However, from a modelling point of view, only the parameter \(\sigma :=\sigma _5\) plays a role. Hence, we fix \(\sigma _4=0.01\) and \(\sigma _6=1\). The coefficients *A*, *B*, *a*, *b*, \(v_0\), \(v_{{\mathrm{max}}}\) and *r* have been experimentally recorded; see, e.g. Jansen et al. (1993), Jansen and Rit (1995) and van Rotterdam et al. (1982). Thus, we fix them according to these values reported, for example, in Table 1 of Ableidinger et al. (2017). In contrast, the connectivity parameter *C*, which represents the average number of synapses between the neural subpopulations and controls to what extent the main population interacts with the interneurons, varies under different physiological constraints. Changing *C* allows, for example, a transition from \(\alpha \)-rhythmic activity to epileptic spiking behaviour; see, e.g. Ableidinger et al. (2017). Here, we focus on the \(\alpha \)-rhythmic activity, which has a frequency of around 10 Hz. Since the parameters \(\sigma \) and \(\mu \) are new in the SDE version (25), they have not yet been estimated. They can be interpreted as stochastic and deterministic external inputs coming from neighbouring or more distant cortical columns, respectively. Thus, together with the internal connectivity parameter *C*, they are of specific interest. Before inferring \(\theta =(\sigma ,\mu ,C)\), we take into account the coefficients *A* and *B* to discuss a model-specific issue of identifiability.

#### 5.2.1 Identifiability issues: the detection of an invariant manifold, i.e. a set of parameters yielding the same type of data

*A*,

*B*and

*C*yield the same type of output, namely \(\alpha \)-rhythmic brain activity. Applying the proposed spectral density-based and measure-preserving ABC Algorithm 1 (ii) for the inference of \(\theta =(A,B,C)\), with given \(\mu =220\) and \(\sigma =2000\), we confirm that the same nonidentifiability arises for the SDE version (25). We choose \(M=30\) observed paths generated assuming

*w*in (7) according to the procedure introduced in Sect. 2.2 (based on \(L=10^5\) iterations) and fix the tolerance level \(\epsilon =0.04{\text {th}}\) percentile to keep \(10^3\) of all the sampled values for \(\theta \), as in the previous examples. Further, we choose independent uniform prior distributions, namely

*A*,

*B*and

*C*are strongly correlated, as observed in the pairwise scatterplots (middle panels) and in the 3-dimensional scatterplot (two different views, lower panels). The cuboid covers all possible values for \(\theta \) drawn from the prior. After running the ABC algorithm, the kept values of \(\theta \) from the ABC posterior form an invariant manifold, in the sense that all the parameter values \(\theta \) lying on this manifold yield similar paths \({\tilde{y}}_\theta \) of the output process. This is shown in Fig. 7, where we report four trajectories that have been simulated with the same pseudo-random numbers but using the parameters \(\theta ^t\) (green dot in Fig. 6) and three of the kept ABC posterior samples lying on the invariant manifold (red, orange and grey dots in Fig. 6). A segment of \(T=10\) is split in the top and middle panels. In addition, we visualise the corresponding estimated invariant densities (bottom left) and invariant spectral densities (bottom right). This explains why the parameters

*A*,

*B*and

*C*are not simultaneously identifiable from the observed data. Similar results are obtained when choosing smaller values of \(\epsilon \). Interestingly, when increasing \(\epsilon \), a second invariant manifold arises. Values for \(\theta \) lying on this manifold yield similar estimated densities and spectral densities that slightly deviate from those derived under the observed data (cf. Section 3 of the supplementary material).

Since the internal connectivity parameter *C* has an important neuronal meaning, in the following we assume *A* and *B* to be known and infer \(\theta =(\sigma ,\mu ,C)\). The estimation of \(\theta =(\sigma ,\mu )\) when *C* is known is reported in the supplementary material.

#### 5.2.2 Inference of \(\theta =(\sigma ,\mu ,C)\)

*C*is known (cf. Figure 21 of the supplementary material). This results from a weak correlation that we detect among the kept ABC posterior samples of the parameters \(\sigma \) and

*C*(figures not reported). The posterior means are equal to

Similar satisfactory results are obtained even when adding a fourth parameter, for example, when inferring \(\theta =(\sigma ,\mu ,C,b)\) (cf. Figure 22 of the supplementary material). When applying Algorithm 1 (ii) to real EEG data (cf. Figure 23 of the supplementary material), the marginal posterior for *b* is centred around the value \(b=50\), which is that reported in the literature. Due to the existence of underlying invariant manifolds, identifiability issues, similar to those reported in Fig. 6, arise when adding further or other coefficients, revealing model-specific issues for the stochastic JR-NMM.

### 5.3 Parameter inference from real EEG data

^{1}Figure 10 shows the first 20 seconds of one of the observed EEG datasets. Here, we simulate \(N=5 \cdot 10^6\) synthetic paths from the output process of the stochastic JR-NMM (25) over the same time interval

*T*as the real data, with a time step \(\varDelta =2 \cdot 10^{-3}\) and \(\epsilon =0.02{\text {nd}}\) percentile. We choose independent uniform priors \(\pi (\theta _j)\) according to

*C*is centred around \(C=135\) that is the reference literature value for \(\alpha \)-rhythmic EEG data.

In Fig. 12 (top two panels), we report the median (blue solid line) and the shaded 95% credible bands (obtained from the central posterior intervals) of the first 10 seconds of trajectories simulated from the fitted stochastic JR-NMM (25). The paths are generated with the numerical splitting scheme (17) for \(\varDelta =2 \cdot 10^{-3}\) and \(T=23.6\), using the \(10^3\) kept ABC posterior samples derived from Algorithm 1 (ii) under the same seed for pseudo-random numbers. The narrow 95% confidence bands suggest how the kept ABC posterior samples yield similar paths. When using different pseudo-random numbers, both the median and the 95% credible bands of the generated trajectories look constant in time, as expected due to the underlying invariant distribution of the model (figures not shown). The bands show a similar oscillatory behaviour as shown in Fig. 10, with an approximate frequency of 10 Hz, successfully reproducing the underlying \(\alpha \)-rhythmic behaviour. This can be clearly distinguished in the middle lower panel, where we report a zoom of 1 second. The successful ABC inference is also confirmed by noting the matches between the invariant densities (bottom left) and the invariant spectral densities (bottom right) estimated from the EEG recording shown in Fig. 10 (red dashed lines) and from the fitted model when considering the median (blue solid lines) and the 95% credible bands (grey shaded areas). The match is poor only for low frequencies of the invariant spectral density, even when choosing broader priors. This may result from a lack of fit of the JR-NMM or of stationarity in the considered EEG data. A deeper investigation of the model (including adapted versions, see, e.g. Wendling et al. 2002) and of its ability in reproducing real EEG data is currently under investigation, but it is out of the scope of this work.

## 6 Conclusion

When performing parameter inference through ABC, crucial and non-trivial tasks are to propose suitable summary statistics and distances to compare the observed and the synthetic datasets. When the underlying models are stochastic, repeated simulations from the same parameter setting yield different outputs, making the comparison between the observed and the synthetic data more difficult. To derive summary statistics that are less sensitive to the intrinsic randomness of the stochastic model, we propose to map the data to their invariant density and invariant spectral density, estimated by a kernel density estimator and a smoothed periodogram, respectively. By doing this, different trajectories of the output process are mapped to the same objects only when they are generated from the same underlying parameters, provided that all parameters are simultaneously identifiable. These transformations are based on the existence of an underlying invariant measure for the model, fully characterised by the parameters. A necessary condition of ABC, and of all other simulation-based methods, is the ability to generate data from the model. This is often taken for granted but, in general, it is not the case. Indeed, exact simulation is rarely possible and property-preserving numerical methods have to be derived.

The combination of the measure-preserving numerical splitting schemes and the use of the spectral density-based distances in the ABC algorithm lead to a successful inference of the parameters, as illustrated on stochastic Hamiltonian type equations when the process \(\mathbf{Y }_\theta \) is observed without measurement error. We validated the proposed ABC approach on both linear model problems, allowing for an exact simulation of the synthetic data, and nonlinear problems, including an application to real EEG data. Our choice of the crucial ingredients (summary statistics and distances based on the underlying invariant distribution and a measure-preserving numerical method) yields excellent results even when applied to ABC in its basic acceptance–rejection form. However, they can be directly applied to more advanced ABC algorithms. In contrast, the ABC method based on the Euler–Maruyama scheme drastically fails. Its performance may improve for “small enough” time steps. However, there is a trade-off between the runtime and the acceptance performance of Algorithm 1 (iii). Indeed, the simulation of one trajectory with a time step \(10^{-4}\) requires approximately hundred times more than the generation of one trajectory using a time step \(10^{-2}\). Hence, a runtime of a few hours would turn to months. In addition, even for “arbitrary small” time steps, one cannot guarantee that the Euler–Maruyama scheme preserves the underlying invariant measure. For these reasons, it is crucial to base our ABC method on the reliable measure-preserving numerical splitting scheme combined with the invariant measure-based distances. Our results were discussed in the case of an observable 1-dimensional output process. However, the approach can be directly applied to *d*-dimensional output processes, \(d>1\), as long as the underlying SDEs are characterised by an invariant distribution and a measure-preserving numerical method can be derived. In particular, one can compute the distances in (8) for each of the *d* components and derive a global distance by combining them, e.g. via their sum. Moreover, to account for possible dependences between the observed components, one can incorporate the cross-spectral densities which are expected to provide further information resulting in an improvement in the performance of the method. An investigation in this direction is currently undergoing. Finally, our proposed ABC method may be also used to investigate invariant manifolds characterised by sets of parameters yielding the same type of data, as illustrated on the stochastic JR-NMM. This may result in a better understanding of the qualitative behaviour of the underlying model and its ability of reproducing the true features of the modelled phenomenon.

## 7 Supplementary material

Further illustrations of the proposed ABC method are available in the provided supplementary material. In particular, we illustrate the performance of: (a) Algorithm 1 (i) for the estimation of the parameters of the critically damped stochastic oscillator, for which the exact simulation is possible; Algorithm 1 (i) applied to the critically and weakly damped stochastic harmonic oscillators (see Sect. 4) when estimating a smaller number of parameters; (b) Algorithm 1 (ii) for the estimation of the parameters of a nonlinear damped stochastic oscillator; (c) Algorithm 1 (ii) for the estimation of the two parameters \(\theta =(\sigma ,\mu )\) of the stochastic JR-NMM, which are of specific interest; (d) Algorithm 1 (ii) for the estimation of \(\theta =(\sigma ,\mu ,C,b)\) of the stochastic JR-NMM, based on simulated and real EEG data. Moreover, we provide an investigation of the influence of \(\epsilon \) on the identifiability issues discussed in Sect. 5.2.1.

## Footnotes

- 1.
The data are available on: http://ntsa.upf.edu/downloads/andrzejak-rg-et-al-2001-indications-nonlinear-deterministic-and-finite-dimensional.

## Notes

### Acknowledgements

Open access funding provided by Austrian Science Fund (FWF). This research was partially supported by the Austrian Science Fund (FWF): W1214-N15, Project DK14. The authors would like to thank the reviewers for their careful and thorough reading and their constructive comments, which helped to improve earlier versions of the manuscript. Their support is highly appreciated.

## Supplementary material

## References

- Ableidinger, M., Buckwar, E.: Splitting Integrators for the stochastic Landau–Lifshitz equation. SIAM J. Sci. Comput.
**38**, A1788–A1806 (2016) MathSciNetzbMATHCrossRefGoogle Scholar - Ableidinger, M., Buckwar, E., Hinterleitner, H.: A Stochastic Version of the Jansen and Rit neural mass model: analysis and numerics. J. Math. Neurosci. (2017). https://doi.org/10.1186/s13408-017-0046-4
- Andrzejak, R.G., Lehnertz, K., Mormann, F., Rieke, C., David, P., Elger, C.E.: Indications of nonlinear deterministic and finite-dimensional structures in time series of brain electrical activity: dependence on recording region and brain state. Phys. Rev. E
**64**, 061907 (2001)CrossRefGoogle Scholar - Arnold, L.: Stochastic Differential Equations: Theory and Applications. Wiley, New York (1974)zbMATHGoogle Scholar
- Barber, S., Voss, J., Webster, M.: The rate of convergence for approximate Bayesian computation. Electron. J. Stat.
**9**(1), 80–105 (2015)MathSciNetzbMATHCrossRefGoogle Scholar - Barnes, C., Filippi, S., Stumpf, M., Thorne, T.: Considerate approaches to constructing summary statistics for ABC model selection. Stat. Comput.
**22**(6), 1181–1197 (2012)MathSciNetzbMATHCrossRefGoogle Scholar - Beaumont, M.A., Zhang, W., Balding, D.J.: Approximate Bayesian computation in population genetics. Genetics
**162**(4), 2025–2035 (2002) Google Scholar - Bernton, E., Jacob, P.E., Gerber, M., Robert, C.P.: Approximate Bayesian computation with the Wasserstein distance. J. R. Stat. Soc. B
**81**(2), 235–269 (2019)MathSciNetzbMATHCrossRefGoogle Scholar - Biau, G., Cérou, F., Guyader, A.: New Insights Into approximate Bayesian computation. Ann. I. H. Poincare B
**51**(1), 376–403 (2015)MathSciNetzbMATHCrossRefGoogle Scholar - Blanes, S., Casas, F., Murua, A.: Splitting and composition methods in the numerical integration of differential equations. Bol. Soc. Esp. Mat. Apl.
**45**, 89–145 (2009)zbMATHGoogle Scholar - Blum, M.G.B.: Approximate Bayesian computation: a nonparametric perspective. J. Am. Stat. Assoc.
**105**(491), 1178–1187 (2010a)MathSciNetzbMATHCrossRefGoogle Scholar - Blum, M.G.B.: Choosing the summary statistics and the acceptance rate in approximate Bayesian computation. In: Lechevallier, Y., Saporta, G. (eds.) Proceedings of COMPSTAT 2010, pp 47–56. Physica-Verlag HD, Heidelberg (2010b)CrossRefGoogle Scholar
- Boys, R.J., Wilkinson, D.J., Kirkwood, T.B.L.: Bayesian inference for a discretely observed stochastic kinetic model. Stat. Comput.
**18**, 125–135 (2008)MathSciNetCrossRefGoogle Scholar - Bréhier, C.E., Goudenège, L.: Analysis of Some splitting schemes for the stochastic Allen–Cahn equation. Discrete Cont. Dyn. B
**24**, 4169–4190 (2019)MathSciNetzbMATHGoogle Scholar - Cadonna, A., Kottas, A., Prado, R.: Bayesian mixture modeling for spectral density estimation. Stat. Probab. Lett.
**125**, 189–195 (2017)MathSciNetzbMATHCrossRefGoogle Scholar - Cornuet, J.M., Santos, F., Beaumont, M., Robert, C.P., Marin, J., Balding, D.J., Guillemaud, T., Estoup, A.: Inferring population history with DIY ABC: a user-friendly approach to approximate Bayesian computation. Bioinformatics
**24**(23), 2713–2719 (2008)CrossRefGoogle Scholar - Drovandi, C.C., Pettitt, A.N., McCutchan, R.: Exact and approximate Bayesian inference for low integer-valued time series models with intractable likelihoods. Bayesian Anal.
**11**, 325–352 (2016)MathSciNetzbMATHCrossRefGoogle Scholar - Eddelbuettel, D., François, R.: Rcpp: Seamless R and C++ integration. J. Stat. Softw.
**40**(8), 1–18 (2011)CrossRefGoogle Scholar - Fan, Y., Sisson, S.A.: ABC samplers. In: Sisson, S.A., Fan, Y., Beaumont, M. (eds.) Handbook of Approximate Bayesian Computation, pp. 87–123. CRC Press, Taylor & Francis Group: chap 4 (2018)Google Scholar
- Fearnhead, P., Prangle, D.: Constructing summary statistics for approximate Bayesian computation: semi-automatic approximate Bayesian computation. J. R. Stat. Soc. B
**74**(3), 419–474 (2012)MathSciNetzbMATHCrossRefGoogle Scholar - Jansen, B.H., Rit, V.G.: Electroencephalogram and visual evoked potential generation in a mathematical model of coupled cortical columns. Biol. Cybern.
**73**(4), 357–366 (1995)zbMATHCrossRefGoogle Scholar - Jansen, B.H., Zouridakis, G., Brandt, M.E.: A neurophysiologically-based mathematical model of flash visual evoked potentials. Biol. Cybern.
**68**, 275–283 (1993)CrossRefGoogle Scholar - Jasra, A.: Approximate Bayesian computation for a class of time series models. Int. Stat. Rev.
**83**(3), 405–435 (2015)MathSciNetCrossRefGoogle Scholar - Jiang, B., Wu, T.-Y., Zheng, C., Wong, W.H.: Learning summary statistics for approximate Bayesian computation via deep neural network. Stat. Sinica
**27**(4), 1595–1618 (2017)MathSciNetzbMATHGoogle Scholar - Kloeden, P.E., Platen, E.: Numerical Solution of Stochastic Differential Equations. Springer, Berlin (1992)zbMATHCrossRefGoogle Scholar
- Kypraios, T., Neal, P., Prangle, D.: A tutorial introduction to Bayesian inference for stochastic epidemic models using approximate Bayesian computation. Math. Biosci.
**287**, 42–53 (2017)MathSciNetzbMATHCrossRefGoogle Scholar - Leimkuhler, B., Matthews, C.: Molecular Dynamics: With Deterministic and Stochastic Numerical Methods. Springer, Cham (2015)zbMATHGoogle Scholar
- Leimkuhler, B., Matthews, C., Stoltz, G.: The computation of averages from equilibrium and nonequilibrium Langevin molecular dynamics. IMA J. Numer. Anal.
**36**(1), 16–79 (2016)MathSciNetzbMATHGoogle Scholar - Lintusaari, J., Gutmann, M., Dutta, R., Kaski, S., Corander, J.: Fundamentals and recent developments in approximate Bayesian computation. Syst. Biol.
**66**(1), e66–e82 (2017) Google Scholar - Malham, S.J., Wiese, A.: Chi-square simulation of the CIR process and the Heston model. Int. J. Theor. Appl. Finance
**16**(3), 1350014 (2013)MathSciNetzbMATHCrossRefGoogle Scholar - Marin, J.-M., Pudlo, P., Robert, C.P., Ryder, R.: Approximate Bayesian computational methods. Stat. Comput.
**22**(6), 1167–1180 (2012)MathSciNetzbMATHCrossRefGoogle Scholar - Martin, G.M., McCabe, B.P.M., Frazier, D.T., Maneesoonthorn, W., Robert, C.P.: Auxiliary likelihood-based approximate Bayesian computation in state space models. J. Comput. Graph. Stat.
**28**(3), 1–31 (2019)MathSciNetCrossRefGoogle Scholar - Mattingly, J.C., Stuart, A.M., Higham, D.J.: Ergodicity for SDEs and approximations: locally Lipschitz vector fields and degenerate noise. Stoch. Process. Appl.
**101**(2), 185–232 (2002)MathSciNetzbMATHCrossRefGoogle Scholar - Maybank, P., Bojak, I., Everitt, R.: Fast approximate Bayesian inference for stable differential equation models (2017). arXiv.org/abs/1706.00689
- McKinley, T.J., Vernon, I., Andrianakis, I., McCreesh, N., Oakley, J., Nsubuga, R., Goldstein, M., White, R.: Approximate Bayesian computation and simulation-based inference for complex Stochastic epidemic models. Stat. Sci.
**33**(1), 4–18 (2017)MathSciNetzbMATHCrossRefGoogle Scholar - Mclachlan, R., Quispel, G.: Splitting methods. Acta Numer.
**11**, 341–434 (2002)MathSciNetzbMATHCrossRefGoogle Scholar - Milstein, G.N., Tretyakov, M.V.: Stochastic Numerics for Mathematical Physics. Scientific Computation. Springer, Berlin (2004)zbMATHCrossRefGoogle Scholar
- Misawa, T.: A Lie algebraic approach to numerical Integration of Stochastic differential equations. SIAM J. Sci. Comput.
**23**(3), 866–890 (2001)MathSciNetzbMATHCrossRefGoogle Scholar - Moores, M.T., Drovandi, C.C., Mengersen, K., Robert, C.P.: Pre-processing for approximate Bayesian computation in image analysis. Stat. Comput.
**25**, 23–33 (2015)MathSciNetzbMATHCrossRefGoogle Scholar - Mori, U., Mediburu, A., Lozano, J.A.: Distance measures for time series in R: the TSdist package. R Journal
**8**, 455–463 (2016)CrossRefGoogle Scholar - Moro, E., Schurz, H.: Boundary preserving semianalytic numerical algorithms for Stochastic differential equations. SIAM J. Sci. Comput.
**29**, 1525–1549 (2007)MathSciNetzbMATHCrossRefGoogle Scholar - Muskulus, M., Verduyn-Lunel, S.: Wasserstein distances in the analysis of time series and dynamical systems. Physica D
**240**(1), 45–58 (2011)MathSciNetzbMATHCrossRefGoogle Scholar - Picchini, U.: Inference for SDE models via approximate Bayesian computation. J. Comput. Graph. Stat.
**23**(4), 1080–1100 (2014)MathSciNetCrossRefGoogle Scholar - Picchini, U., Forman, J.L.: Accelerating inference for diffusions observed with measurement error and large sample sizes using approximate Bayesian computation. J. Stat. Comput. Simul.
**86**(1), 195–213 (2016)MathSciNetCrossRefGoogle Scholar - Picchini, U., Samson, A.: Coupling stochastic EM and approximate Bayesian computation for parameter inference in state-space models. Comput. Stat.
**33**(1), 179–212 (2018)MathSciNetzbMATHCrossRefGoogle Scholar - Pons, O.: Functional Estimation for Density. Regression Models and Processes. World Scientific Publishing, Singapore (2011)zbMATHCrossRefGoogle Scholar
- Prangle, D.: Adapting the ABC distance function. Bayesian Anal.
**12**(1), 289–309 (2017) MathSciNetzbMATHCrossRefGoogle Scholar - Prangle, D.: Summary Statistics in approximate Bayesian computation. In: Sisson, S.A., Fan, Y., Beaumont, M. (eds.) Handbook of Approximate Bayesian Computation, pp. 125–152. CRC Press, Taylor & Francis Group: chap 5 (2018)Google Scholar
- Prangle, D., Blum, M.G.B., Popovic, G., Sisson, S.A.: Diagnostic tools for approximate Bayesian computation using the coverage property. Aust. N. Z. J. Stat.
**56**(4), 309–329 (2014)MathSciNetzbMATHCrossRefGoogle Scholar - Quinn, B., Clarkson, I., Mckilliam, R.: On the periodogram estimators of periods from interleaved sparse, noisy timing data. In: 2014 IEEE Workshop on Statistical Signal Processing, pp. 232–235 (2014)Google Scholar
- R Development Core Team: R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (2011)Google Scholar
- Robert, C.P.: Approximate Bayesian computation: a survey on recent results. In: Cools, R., Nuyens, D. (eds.) Monte Carlo and Quasi-Monte Carlo Methods, pp. 185–205. Springer, Cham (2016)CrossRefGoogle Scholar
- Sason, I., Verdú, S.: \(f\)-Divergence Inequalities. IEEE Trans. Inf. Theory
**62**(11), 5973–6006 (2016)MathSciNetzbMATHCrossRefGoogle Scholar - Sisson, S.A., Fan, Y., Beaumont, M.: Handbook of Approximate Bayesian Computation. Chapman & Hall/CRC Handbooks of Modern Statistical Methods. CRC Press, Taylor & Francis Group (2018)Google Scholar
- Strang, G.: On the construction and comparison of Difference schemes. SIAM J. Numer. Anal.
**5**(3), 506–517 (1968)MathSciNetzbMATHCrossRefGoogle Scholar - Strømmen Melbø, A.H., Higham, D.J.: Numerical simulation of a linear stochastic oscillator with additive noise. Appl. Numer. Math.
**51**(1), 89–99 (2004)MathSciNetzbMATHCrossRefGoogle Scholar - Sun, L., Lee, C., Hoeting, J.A.: Parameter inference and model selection in deterministic and stochastic dynamical models via approximate Bayesian computation: modeling a wildlife epidemic. Environmetrics
**26**(7), 451–462 (2015)MathSciNetCrossRefGoogle Scholar - Tancredi, A.: Approximate Bayesian inference for discretely observed continuous-time multi-state models. Biometrics (2019). https://doi.org/10.1111/biom.13019 CrossRefGoogle Scholar
- Toni, T., Welch, D., Strelkowa, N., Ipsen, A., Stumpf, M.P.H.: Approximate Bayesian computation scheme for parameter inference and model selection in dynamical systems. J. R. Soc. Interface
**6**(31), 187–202 (2009)CrossRefGoogle Scholar - van Rotterdam, A., Lopes da Silva, F., van den Ende, J., Viergever, M.A., Hermans, A.: A model of the spatial-temporal characteristics of the alpha rhythm. Bull. Math. Biol.
**44**, 283–305 (1982)MathSciNetzbMATHCrossRefGoogle Scholar - Vo, B.N., Drovandi, C.C., Pettitt, A.N., Simpson, M.J.: Quantifying uncertainty in parameter estimates for stochastic models of collective cell spreading using approximate Bayesian computation. Math. Biosci.
**263**, 133–142 (2015)MathSciNetzbMATHCrossRefGoogle Scholar - Wendling, F., Bellanger, J.J., Bartolomei, F., Chauvel, P.: Relevance of nonlinear lumped-parameter models in the analysis of depth-EEG epileptic signals. Biol. Cybern.
**83**(4), 367–378 (2000)CrossRefGoogle Scholar - Wendling, F., Bartolomei, F., Bellanger, J.J., Chauvel, P.: Epileptic fast activity can be explained by a model of impaired GABAergic dendritic inhibition. Eur. J. Neurosci.
**15**(9), 89–99 (2002)CrossRefGoogle Scholar - Wood, S.N.: Statistical inference for noisy nonlinear ecological dynamic systems. Nature
**466**(7310), 1102 (2010)CrossRefGoogle Scholar - Zhu, W., Marin, J.M., Leisen, F.: A bootstrap likelihood approach to Bayesian computation. Aust. N. Z. J. Stat.
**58**(2), 227–244 (2016)MathSciNetCrossRefGoogle Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.