Estimation of count time series model with varying frequencies: Application to prevalence rate of diseases

College

College of Science

Department/Unit

Mathematics and Statistics Department

Document Type

Archival Material/Manuscript

Publication Date

12-2019

Abstract

In modeling time series data with varying frequencies, variables at higher frequency are commonly aggregated first to coincide with the usually lower frequency of the dependent variable, and in the process, resulting to information loss. A semiparametric count model for time series data with varying frequencies is proposed. High frequency covariates are incorporated into nonparametric functions (without aggregation) to explain behavior of poisson-distributed count response. The contribution of the covariate with same frequency as the response is assumed to be parametric. Simulation studies and real data application show advantages of the model based on the Mean Absolute Deviation (MAD) over a General Additive Model and an Ordinary Poisson regression model especially on covariates with weak or no autocorrelation.

html

Disciplines

Mathematics

Keywords

Time-series analysis; Parametric modeling

Upload File

wf_no

This document is currently not available here.

Share

COinS