Claim Missing Document
Check
Articles

Found 3 Documents
Search
Journal : Media Statistika

CONWAY-MAXWELL POISSON REGRESSION MODELING OF INFANT MORTALITY IN SOUTH SULAWESI Oktaviana, Oktaviana; Sanusi, Wahidah; Aswi, Aswi; Sukarna, Sukarna; Folorunso, Serifat Adedamola
MEDIA STATISTIKA Vol 17, No 1 (2024): Media Statistika
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/medstat.17.1.45-56

Abstract

Overdispersion is a common problem in count data that can lead to inaccurate parameter estimates in Poisson regression models. Quasi-Poisson and negative binomial regressions are often used to address overdispersion but have limitations, especially with small samples. The Conway-Maxwell Poisson (CMP) regression model, an extension of the Poisson distribution, effectively addresses both overdispersion and underdispersion, even with limited data, due to additional parameters that better control data dispersion. The Infant Mortality Rate (IMR) is a critical public health indicator, reflecting healthcare quality and broader social, economic, and environmental factors. Accurate IMR estimation is essential for evaluating health policies. This study aims to (1) identify overdispersion in IMR data from South Sulawesi, (2) model IMR using CMP regression, and (3) identify factors influencing IMR. The dataset includes IMR, Low Birth Weight (LBW), diarrhea, asphyxia, pneumonia, and exclusive breastfeeding. Analysis showed significant overdispersion with a ratio of 4.639, making CMP the optimal model with an AIC of 186.845. Significant factors identified were LBW, asphyxia, pneumonia, and exclusive breastfeeding. These findings advance statistical methodologies for count data analysis and offer a more accurate approach to evaluating public health policies, supporting efforts to reduce infant mortality in South Sulawesi Province.
MAKING BAYESIAN DISEASE MAPPING EASY AND INTERACTIVE: AN R SHINY APPLICATION Aswi, Aswi; Tiro, Muhammad Arif; Sudarmin, Sudarmin; Sukarna, Sukarna; Awi, Awi; Nurwan, Nurwan; Cramb, Susanna
MEDIA STATISTIKA Vol 16, No 2 (2023): Media Statistika
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/medstat.16.2.148-159

Abstract

Spatial analysis of count data is important in epidemiology and other domains to identify spatial patterns. While Bayesian spatial models are a popular approach, they do require detailed knowledge of the process for model fitting, checking, and visualising results. Although a number of R packages are available to simplify running the model, there are still complexities when checking the model. This paper aims to provide a user-friendly and interactive R Shiny web application for the analysis of spatial data using Bayesian spatial Conditional Autoregressive Leroux models. The web application is built with the integration of the R packages shiny and CARBayes. The required data are the number of cases, population, and optionally some covariates for each region. In this case, we used Covid-19 data in 2021 in South Sulawesi province, Indonesia. This application enables fitting a Bayesian spatial CAR Leroux model under several hyperpriors and selecting the most appropriate through comparing several goodness of fit measures. The application also enables checking convergence, plus obtaining and visualising in an interactive map the relative risk of disease for each region.
COMPARISON OF LOGISTIC MODEL TREE AND RANDOM FOREST ON CLASSIFICATION FOR POVERTY IN INDONESIA Sukarna, Sukarna; Notodiputro, Khairil Anwar; Sartono, Bagus
MEDIA STATISTIKA Vol 16, No 2 (2023): Media Statistika
Publisher : Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.14710/medstat.16.2.112-123

Abstract

Classification methods are commonly employed to ensure homogeneous data within each group, facilitating the prediction of specific categories. The most frequently used classification models are Logistic Model Tree (LMT) and Random Forest (RF). This study aims to assess the accuracy rate in predicting the poverty status of regencies or towns across Indonesia, utilizing eight independent variables. The entire dataset was obtained from the official Central Bureau of Statistics website. The study investigates the accuracy of various iterations and combinations of training data. The results indicate that RF outperforms LMT in terms of accuracy, achieving a 100% improvement in iterations k=10 and k=500 and a 75% improvement in iteration k=100. Consequently, the RF proves to be more effective than the LMT for analyzing Indonesian poverty data, especially when incorporating all eight independent variables.