Comparative Analysis of Sarima and Setar Models in Predicting Pneumonia Cases in Kenya

Fredrick Agwata Nyamato,Anthony Wanjoya,Thomas Mageto

doi:10.11648/j.ijdsa.20200601.16

Fredrick Agwata Nyamato, Anthony Wanjoya + Show 1 more

Open Access

https://doi.org/10.11648/j.ijdsa.20200601.16

Copy DOI

Abstract

Kenya is a country located in Eastern part of Africa with approximate population of 46.5 million, with majority of the population constituting youths under the age of 35 years. The country has experienced increased morbidity rate arising from Pneumonia disease like other countries all over the world. As per recent studies 2 million children lose lives from pneumonia disease [1]. This study applies two models, one is linear model Seasonal autoregressive model (SARIMA) and another is a non-linear model called self-Excited Threshold Autoregressive (SETAR) in projection of cases in Kenya. Data for usage for purpose of this study was obtained Ministry of Health of Kenya of a period of 20 years from January 1999 to December 2018. The data collected is seasonal the number of case from period to period depending on climatic condition. Although both models performs well in pneumonia projection, non-linear SETAR models outperforms linear SARIMA. By carrying out a comparative analysis by use of Diebold-Mariano test, which revealed that there were no significant difference in the forecasting performance of the two models. The best model identified between the two models i.e. SETAR which best fit the data, can be applied in predicting pneumonia cases beyond the period under consideration. Other studies can be carried to come up with a model for every specific region in the country, to assist in resources allocation to specific parts of the country.

Highlights

Pneumonia is characterized by an acute infection of the lungs, which results in coughing, fever, chills, muscle aches, and difficulty in breathing to its victims
The World Health Organization (WHO) [3] estimated that there were more than 150 million cases of pneumonia each year and killing 1.6 million which accounts for 19% of all deaths worldwide
After we have found that self-Excited Threshold Autoregressive (SETAR) (2; 4, 3) model with threshold variable ) as the best model that fits the data well since it has the minimum value for Akaike Information Criterion (AIC)

Summary

Introduction

Pneumonia is characterized by an acute infection of the lungs, which results in coughing, fever, chills, muscle aches, and difficulty in breathing to its victims. In the year 2011, there were an estimated 120 million episodes of childhood pneumonia globally of which 14 million progressed to severe disease with 1.3 million deaths [2]. Most of these deaths (81%) occurred in children under 2 years of age. The primary causative agent of Pneumonia is known by its scientific name Streptococcus pneumonia, through understanding this causative agent and reviewing strategies which have been deployed to manage the disease on a global scale will be important in reducing deaths caused it Through this effort, the world will achieve easy access and efficiency in its treatment and eventually reduce its detrimental effects.

Methods

Results

Conclusion