Seasonality of Influenza: An Investigation Using Time Series, Spectral Analysis & Epidemiological Models
Loading...
Authors
Boteler, Claire
Date
Type
thesis
Language
eng
Keyword
Time Series , Spectral Analysis , Influenza , Susceptible-Infected Model , Harmonic Analysis
Alternative Title
Abstract
Influenza is an infectious disease. Its seasonality is commonly modeled focusing on the yearly cycle. The log transform of pneumonia and influenza mortality data from the United States from 1910 to 2015 showed an approximate 30 year recurring pattern. The data were analyzed in the time domain and then in the frequency domain using the multitaper spectrum estimation method. Significant frequencies were identified using the F-test and confirmed by the presence of distinct square peaks in the spectrum plot. This monthly sampled data identified long-term trends at 136 year, 14 year, 11 year and 7.5 year periods as well as high frequencies of 2, 3, 4 and 5 c/y. These intriguing results led to the analysis of influenza incidence data from the United States, United Kingdom, Australia and Japan. Simulated data was generated from a susceptible-infected epidemiological model with the contact rate of influenza transmission set to a one-year period. It was also modified to include multiple periodic terms that were identified as significant in the collected datasets. After comparing the spectral results of the collected and simulated data, similarities were easier to notice and interpret. Spectral analysis identifies frequencies that contribute to the variations in the data. These frequencies could have individual biological significance and/or aid in the attempt to represent the data that are non-sinusoidal by summing their sinusoids. The high frequencies are most likely harmonics of the yearly cycle, where 2, 8 and 9 c/y are recommended to be considered in inclusion in multi-term periodic models. Simulations presented the opportunity to compare the effect of sampling rates and data length on the spectral results. It is recommended to use weekly sampled data of at least 30 years. The results from investigating influenza morbidity and mortality data in the frequency domain aid to better understand the patterns of the data, which can be used to improve forecasting of influenza data.
Description
Citation
Publisher
License
Attribution 3.0 United States
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.