Integer-valued Time Series Model via Generalized Linear Models Technique of Estimation

Authors

  • Rasaki Olawale Olanrewaju Department of Statistics, Distance Learning Center (DLC), University of Ibadan, Nigeria

DOI:

https://doi.org/10.21467/ias.4.1.35-43

Abstract

The paper authenticated the need for separate positive integer time series model(s). This was done from the standpoint of a proposal for both mixtures of continuous and discrete time series models. Positive integer time series data are time series data subjected to a number of events per constant interval of time that relatedly fits into the analogy of conditional mean and variance which depends on immediate past observations. This includes dependency among observations that can be best described by Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model with Poisson distributed error term due to its positive integer defined range of values. As a result, an integer GARCH model with Poisson distributed error term was formed in this paper and called Integer Generalized Autoregressive Conditional Heteroscedasticity (INGARCH). Iterative Reweighted Least Square (IRLS) parameter estimation technique type of the Generalized Linear Models (GLM) was adopted to estimate parameters of the two spilt models; Linear and Log-linear INGARCH models deduced from the identity link function and logarithmic link function, respectively. This resulted from the log-likelihood function generated from the GLM via the random component that follows a Poisson distribution. A study of monthly successful bids of auction from 2003 to 2015 was carried out. The Probabilistic Integral Transformation (PIT) and scoring rule pinpointed the uniformity of the linear INGARCH than that of the log-linear INGARCH in describing first order autocorrelation, serial dependence and positive conditional effects among covariates based on the immediate past. The linear INGARCH model outperformed the log-linear INGARCH model with (AIC = 10514.47, BIC = 10545.01, QIC = 34128.56) and (AIC = 37588.83, BIC = 37614.28, QIC = 37587.3), respectively.

Keywords:

Generalized Linear Model (GLM), Integer Generalized Autoregressive Conditional Heteroscedasticity (INGARCH), Iterative Reweighted Least Square (IRLS)

Downloads

Download data is not yet available.

References

L. Tobias, K. Fokianos, and R. Fried, “An R Package for Analysis of Count Time Series Following Generalized Linear Models.” Journal of Statistical Software. vol.85, no.5, pp. 1-51, Nov. 2017.

M. Salmon, D. Schumacher, and M. Höhle, “Monitoring Count Time Series in R: Aberration Detection in Public Health Surveillance.” Journal of Statistical Software, vol. 70, no.10, pp. 1–35, April 2016.

M. S. Christopher, L.B. Robert, L. M. Kerrie, and J. Robert, “PySSM: A Python Module for Bayesian Inference of Linear Gaussian State Space Models”, Journal of Statistical Software, vol. 57, no. 6, Mar. 2014.

A. Agosto, G. Cavaliere, D. Kristensen, and A. Rahbek, “Modeling Corporate Defaults: Poisson Autoregressions with Exogenous Covariates (PARX)”. CREATES Research Paper School of Economics and Management, University of Aarhus, 2015.

V. Christou, and K. Fokianos, “On Count Time Series Prediction”. Journal of Statistical Computation and Simulation, vol. 85, no.2, pp. 357–373, 2015.

H. Elsaied, and R. Fried, “Robust Fitting of INARCH Models”. Journal of Time Series Analysis, vol.35, no.6, pp. 517–535, Jun. 2014.

R. Jung, and A. Tremayne, “Useful Models for Time Series of Counts or Simply Wrong Ones?” Advances in Statistical Analysis, vol. 95, no.1, pp. 59–91, 2011.

K. Fokianos, “Count Time Series Models”. In T. Subba Rao, S Subba Rao, CR Rao (eds.), Time Series – Methods and Applications, Handbook of Statistics, pp. 315–347, 2012. Elsevier, Amsterdam.

D. Tjøstheim, “Count Time Series with Observation-Driven Autoregressive Parameter Dynamics.” In R Davis, S Holan, R Lund, N Ravishanker (eds.), Handbook of Discrete Valued Time Series, Handbooks of Modern Statistical Methods, Chapman & Hall, London, pp. 77–100, 2015.

M. Yang, J.E. Cavanaugh, and G.K.D. Zamba, “State-Space Models for Count Time Series with Excess Zeros.” Statistical Modeling, vol.15, no.1, pp. 70–90, Nov. 2015.

P. Brockwell, and R. Davis, “Introduction to Time Series and Forecasting”, 2nd edition. Springer-Verlag, New York, 2010.

J. Durbin, and S. Koopman, “Time Series Analysis by State Space Methods.” Oxford University Press, Oxford, May 2012.

K. Fokianos, and D. Tjøstheim, “Nonlinear Poisson Auto regression.” The Annals of the Institute of Statistical Mathematics, vol. 64, no.6, pp. 1205–1225, Oct. 2012.

K. Fokianos, “Statistical Analysis of Count Time Series Models: A GLM Perspective.” In R Davis, S Holan, R Lund, N Ravishanker (eds.), Handbook of Discrete-Valued Time Series, Handbooks of Modern Statistical Methods, Chapman & Hall, London, pp. 3–28, May 2015.

J.A. Nelder, and R.W.M. Wedderburn, “Generalized Linear Models”. Journal of the Royal Statistical Society, vol.135, no. 3, pp.370–384, 1972.

J. Fox, “Maximum-Likelihood Estimation of Generalized Linear Models: Applied Regression Analysis and Generalized Linear Models” (2nd Edition). New York: Springer 2008.

Downloads

Published

2018-04-29

Issue

Section

Research Article

How to Cite

[1]
R. O. Olanrewaju, “Integer-valued Time Series Model via Generalized Linear Models Technique of Estimation”, Int. Ann. Sci., vol. 4, no. 1, pp. 35–43, Apr. 2018.