<b>KFAS</b>: Exponential Family State Space Models in <i>R</i>

Helske, Jouni

doi:10.18637/jss.v078.i10

Cited by 92 publications

(73 citation statements)

References 26 publications

(40 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The Poisson state space model is solved by using the methodology applied to the exponential family of state space models. The algorithm used here is the KFAS (Kalman Filter and Smoother for Exponential Family State Space Models) algorithm developed by Helske [2015] and available at http://cran.r-project.org/ package5KFAS. The Poisson model is approximated by a Gaussian model by replacing the observations by pseudoobservations using the log-link function # t 5log k t ð Þ.…”

Section: Space State Model Of Extremes Occurrencementioning

confidence: 99%

Long‐term variations in global sea level extremes

et al. 2015

View full text Add to dashboard Cite

Decadal to multidecadal variations in sea level extremes unrelated to mean sea level changes have been investigated using long tide gauge records distributed worldwide. A state space approach has been applied that provides robust solutions and uncertainties of the time evolving characteristics of extremes, allowing for data gaps and uneven sampling, both common features of historical sea level time series. Two different models have been formulated for the intensity and for the occurrence of extreme sea level events and have been applied independently to each tide gauge record. Our results reveal two key findings: first, the intensity and the frequency of occurrence of extreme sea levels unrelated to mean sea level vary coherently on decadal scales in most of the sites examined (63 out of 77) and, second, extreme sea level changes are regionally consistent, thus pointing toward a common large‐scale forcing. This variability of extremes associated with climate drivers should be considered in the framework of climate change studies.

show abstract

Section: Space State Model Of Extremes Occurrencementioning

confidence: 99%

Long‐term variations in global sea level extremes

et al. 2015

View full text Add to dashboard Cite

show abstract

“…The LR model is implemented in the glm R package and the KF model is implemented in the KFAS R package (Helske, 2017). The KF model approximates the posterior density by simulating 1000 trials over a section of a single test set of size 108, 035 and is considerably more compute and memory intensive to evaluate than the other classifiers.…”

Section: Resultsmentioning

confidence: 99%

Sequence classification of the limit order book using recurrent neural networks

Dixon

2018

Journal of Computational Science

View full text Add to dashboard Cite

Recurrent neural networks (RNNs) are types of artificial neural networks (ANNs) that are well suited to forecasting and sequence classification. They have been applied extensively to forecasting univariate financial time series, however their application to high frequency trading has not been previously considered. This paper solves a sequence classification problem in which a short sequence of observations of limit order book depths and market orders is used to predict a next event price-flip. The capability to adjust quotes according to this prediction reduces the likelihood of adverse price selection. Our results demonstrate the ability of the RNN to capture the non-linear relationship between the near-term price-flips and a spatio-temporal representation of the limit order book. The RNN compares favorably with other classifiers, including a linear Kalman filter, using S&P500 E-mini futures level II data over the month of August 2016. Further results assess the effect of retraining the RNN daily and the sensitivity of the performance to trade latency.

show abstract

“…Even though dynr can specify some models that other programs cannot, all of the features of other programs that exist for time series modeling are not subsets of dynr. For example, KFAS allows for nonlinear measurement (Helske, 2017a) which is not currently possible in dynr. Moreover, SsfPack has nonlinear measurement capabilities along with many MCMC methods that dynr lacks (Koopman et al, 1999).…”

Section: Discussionmentioning

confidence: 99%

“…However, pomp does not currently support regime-switching functionality beyond the regime switching found in hidden Markov modeling. Helske (2017a) included a review of numerous other packages for non-Gaussian time series models which generally do not involve latent variables.…”

Section: Introductionmentioning

confidence: 99%

Whats for dynr: A Package for Linear and Nonlinear Dynamic Modeling in R

Ou¹,

Hunter²,

Chow³

2019

The R Journal

View full text Add to dashboard Cite

Intensive longitudinal data in the behavioral sciences are often noisy, multivariate in nature, and may involve multiple units undergoing regime switches by showing discontinuities interspersed with continuous dynamics. Despite increasing interest in using linear and nonlinear differential/difference equation models with regime switches, there has been a scarcity of software packages that are fast and freely accessible. We have created an R package called dynr that can handle a broad class of linear and nonlinear discrete-and continuous-time models, with regime-switching properties and linear Gaussian measurement functions, in C, while maintaining simple and easy-tolearn model specification functions in R. We present the mathematical and computational bases used by the dynr R package, and present two illustrative examples to demonstrate the unique features of dynr.

show abstract

KFAS: Exponential Family State Space Models in R

Cited by 92 publications

References 26 publications

Long‐term variations in global sea level extremes

Long‐term variations in global sea level extremes

Sequence classification of the limit order book using recurrent neural networks

Whats for dynr: A Package for Linear and Nonlinear Dynamic Modeling in R

Contact Info

Product

Resources

About