Events

The webinar "STAT-SHOW #9 Pemodelan Data Longitudinal dan Spatio-Temporal" commenced with welcoming remarks, setting the stage for a discussion on the advancements in statistics and data science, aimed at fostering collaboration within the Indonesian academic and professional community. The moderator introduced the distinguished speakers: Prof. Dr. Ir. Muhammad Nur Aidi from IPB University and Ibu Aswi from Universitas Negeri Makassar. Prof. Nur Aidi initiated the technical presentations by elucidating the concept of spatio-temporal longitudinal data modeling, which integrates temporal and spatial dimensions to analyze data observed repeatedly over time across different geographical locations. He provided a detailed explanation of longitudinal data, emphasizing its role in understanding patterns of change through repeated observations, using examples like child growth and economic trends. He then delved into the Linear Mixed-Effects Model (LMME), illustrating its application with and without predictors, and explained the estimation and interpretation of fixed and random effects using student scores over semesters as a practical example. Prof. Nur Aidi further discussed the advantages of LMME in handling hierarchical data and separating variability, while also addressing its challenges, such as determining the random effects structure and sensitivity to normality assumption violations. Transitioning to spatio-temporal models, he introduced techniques like spatio-temporal kriging and Space-Time Autoregressive (STAR) models, providing examples of their calculation and application, including the estimation of values at unobserved locations and times. Finally, Prof. Nur Aidi synthesized these concepts by explaining longitudinal spatio-temporal models, outlining their components, and demonstrating their use in modeling air pollution across different cities over several years, thereby highlighting their capability to enhance the understanding of data dynamics and improve prediction accuracy. Following Prof. Nur Aidi, Ibu Aswi shared her extensive research on spatio-temporal models in the context of epidemiological research, focusing on areal data and the application of the Standardized Incidence Ratio (SIR) for assessing disease risk in small areas. She mentioned various Bayesian models and software packages like CARBayes and R-INLA, emphasizing how spatio-temporal models can capture both spatial and temporal variations in disease risk. Ibu Aswi presented a study evaluating the impact of a small number of areas on spatial estimation using simulated data and compared the performance of different Bayesian models. She further discussed her work on the spatio-temporal analysis of dengue fever in Makassar, Indonesia, comparing several Bayesian spatio-temporal models. Her presentation also included research on the relationship between climate variability and dengue in Makassar, focusing on the role of covariates and spatial priors in forming group structures in dengue risk. Ibu Aswi also shared her research extending Bayesian approaches to survival models, incorporating spatial priors to explain the hospitalization duration of dengue patients, and discussed collaborative research with IPB University on childhood stunting in Indonesia, utilizing Bayesian spatial conditional autoregressive models to identify influencing factors. She concluded by mentioning ongoing research collaborations on tuberculosis and socioeconomic factors affecting stunting in Indonesia, ending her presentation with the insightful quote from George Box, "All models are wrong, but some are useful."

STAT-SHOW #8" is a webinar focusing on LASSO Regression. The webinar aims to share insights in statistics and data science and foster collaboration in Indonesia. The session features two speakers: Dr. Agus Muhammad H. S.Si., M.T. from IPB University, who provides an overview of LASSO Regression, and Dr. Muhamad Yunus, S.Si., M.Si. from UIN Sultan Thaha Saifudin Jambi, who discusses Beta Regression with Group and Sparse Group LASSO. Dr. Agus explains that LASSO Regression is useful for high-dimensional data and multicollinearity issues. It combines subset selection and shrinkage techniques for easier interpretation and model stability, using L1 regularization. He also demonstrates the process using R packages and cross-validation to find the optimal lambda value, and compares its performance with other methods. Dr. Yunus discusses the development of estimation methods for beta regression models using group and sparse group LASSO constraints for responses with values between 0 and 1, like the Village Development Index. Simulations show the effectiveness of these methods, and they are applied to empirical data from Indonesia.

"STAT-SHOW #7" is a webinar focusing on "Sample Design and Ensemble Analysis". This is the seventh webinar in a series that aims to share knowledge and discuss the latest developments in statistics and data science, while also fostering collaboration among academics, practitioners, and students in Indonesia. Dr. Asep Rusiana discusses enhancing the accuracy of variable importance in machine learning using ensemble analysis and simulated annealing. He explains a method to combine multiple variable importance measures into a single metric for better interpretability of complex models. Dr. Rusiana also details the application of simulated annealing, an optimization algorithm, to refine the ensemble variable importance, using a food security dataset as an example. Furthermore, he highlights the advantages of ensemble variable importance over individual models, noting its improved accuracy with uncorrelated predictors.
Anisa provides a foundational understanding of sample design and its crucial role in data analysis. She outlines the traditional sample design process, from defining the target population to selecting appropriate sampling techniques. Anisa also discusses the transition from traditional survey data to big data sources, emphasizing the need for preprocessing in the latter. She illustrates the application of clustering and classification methods to real-world datasets, such as food security and crime rates

"STAT-Show #6" focuses on interpretable machine learning. The event aims to share insights and foster discussions on the latest developments in statistics and data science, while also strengthening connections among academics, practitioners, and students in Indonesia. The event features two speakers: Dr. Dedi Dwi Prasetyo, who discusses the balance between interpretability and accuracy in statistical and machine learning models, and Dr. Bagus Sartono, who focuses on enhancing the utilization of predictive models. The moderator is Herlin Fransiska. Dr. Prasetyo discussed the balance between interpretability and accuracy in machine learning models. He highlighted that while complex models can be highly accurate, they often lack interpretability. He used various statistical models as examples to illustrate this trade-off and emphasized the importance of understanding the fundamentals of mathematics and statistics for effective data science. He also introduced the concept of explainable AI, which aims to make complex models more understandable. Dr. Sartono focused on the use of models for both understanding and prediction. He explained that while machine learning models excel at prediction, they are often less interpretable. He introduced the field of interpretable machine learning, which aims to develop models that are both accurate and understandable. He discussed the importance of interpretability for building trust, ensuring transparency, and improving models. He also touched on different types of interpretability and techniques for making complex models more interpretable.

"STAT-SHOW #5" focuses on the application of the Generalized Linear Mixed Model (GLMM) in analyzing human development data. The session features Dr. Kusman Sad from IPB University and Dr. Etis Sunandi from the University of Bengkulu. Dr. Kusman explains the basics of GLMM, contrasting it with linear models (LM) and generalized linear models (GLM), and highlights its three main components: random component, linear predictor, and link function. He also touches upon parameter estimation methods and the connection between GLMM and Small Area Estimation (SAE) in the context of human development indicators like the Human Development Index (HDI) [15:33], [17:41]. Additionally, he shares research topics and publications from IPB related to HDI and its composite indices. Dr. Etis presents her research on developing SAE models for binary response data, focusing on addressing overdispersion and measurement errors in covariates. She discusses three specific models: CBBHL, CIBB APHL, and SAE ME BBHL. Furthermore, she demonstrates the application of these models using real-world data, such as literacy rates and poverty rates in Bengkulu and East Java provinces.

"STAT-SHOW #3" discusses the applications of machine learning in the fields of food and health. The event is a collaboration between the Indonesian Higher Education Statistics Forum and several universities. Dr. Farid Muhamadendi's presentation focused on refining food access indicators within Indonesia's food security monitoring system. He discussed the inadequacy of the existing system, which compared current prices to a 3-month average, in detecting systemic price increases. His study explored alternative comparison methods, revealing that comparing prices year-over-year was more effective in identifying significant price changes. Consequently, a new methodology was adopted, utilizing fewer commodities (rice, cooking oil, and eggs) and comparing prices annually to improve the detection and response to food price vulnerabilities. Dr. Nur Hilal Asy'ari's presentation centered on the application of data science in precision medicine. She explained that precision medicine aims to tailor treatments based on individual genetic, environmental, and lifestyle variations. Dr. Asy'ari highlighted both the challenges and opportunities associated with integrating diverse data types, including genomic, clinical, and lifestyle data, to enhance disease prediction and treatment strategies. She also provided specific examples of applications, such as genetic risk prediction, drug repurposing, and the utilization of multi-omics data to gain a deeper understanding of disease mechanisms and identify potential drug targets.

"STAT-SHOW #3" discusses the applications of machine learning in the fields of food and health. The event is a collaboration between the Indonesian Higher Education Statistics Forum and several universities. Dr. Farid Muhamadendi's presentation focused on refining food access indicators within Indonesia's food security monitoring system. He discussed the inadequacy of the existing system, which compared current prices to a 3-month average, in detecting systemic price increases. His study explored alternative comparison methods, revealing that comparing prices year-over-year was more effective in identifying significant price changes. Consequently, a new methodology was adopted, utilizing fewer commodities (rice, cooking oil, and eggs) and comparing prices annually to improve the detection and response to food price vulnerabilities. Dr. Nur Hilal Asy'ari's presentation centered on the application of data science in precision medicine. She explained that precision medicine aims to tailor treatments based on individual genetic, environmental, and lifestyle variations. Dr. Asy'ari highlighted both the challenges and opportunities associated with integrating diverse data types, including genomic, clinical, and lifestyle data, to enhance disease prediction and treatment strategies. She also provided specific examples of applications, such as genetic risk prediction, drug repurposing, and the utilization of multi-omics data to gain a deeper understanding of disease mechanisms and identify potential drug targets.

"STAT-SHOW #2" focuses on the use of big data analytics for sustainable development. Dr. Ruliana's presentation centered on the role of big data in supporting sustainable development. She emphasized that data has become a valuable asset, even more so than oil, and outlined the three key dimensions of big data: volume, velocity, and variety. Dr. Ruliana illustrated the ubiquity of big data with examples such as spam filters, biometric logins, and navigation apps. She shared her research on leveraging real-time climate data from satellites to monitor carbon emissions in Makassar and also touched upon the potential of digital economics in reducing carbon emissions. Furthermore, Dr. Ruliana highlighted various applications of big data for sustainable development, including air quality monitoring, stock market analysis, and social media sentiment analysis. She stressed the importance of updating educational curricula to incorporate big data-related topics and demonstrated the use of Google Colab for data collection. Prof. Dr. Anang Kurnia's presentation focused on utilizing satellite imagery and sentiment analysis for sustainable development. He underscored the essential role of statistics and data science in extracting meaningful insights from diverse data sources. Prof. Kurnia detailed the data analysis process, from problem definition to results communication, and discussed the challenges of applying traditional statistical methods to big data, particularly in hypothesis testing and R-squared values. He presented several case studies, including the use of satellite imagery to estimate household numbers and classify rice growth stages. Additionally, he shared research on predicting economic growth through news sentiment analysis, highlighting the application of deep learning models. Prof. Kurnia concluded by discussing the complexities involved in preprocessing satellite images and converting text into sentiment scores.

The Stats Series event aims to share knowledge and experiences in statistics and data science, emphasizing their collaborative nature in addressing future challenges. Statistical thinking is deemed fundamental in the digital age, aiding in data summarization and evolving into predictive modeling with applications in analyzing government impacts and tracking trends. The rise of digital transactions necessitates this statistical mindset for managing rapid data accumulation. The integration of AI and Big Data, particularly machine learning, enhances statistical analysis, improving efficiency in processing large datasets. Key recommendations include understanding data, investing in technology, fostering data literacy, and promoting collaboration between educational institutions and the private sector. Big data plays a crucial role in climate change monitoring through real-time data collection, enabling accurate predictions and the design of evidence-based policies. Machine learning models can predict extreme weather events, and big data enhances transparency in the global carbon market, with blockchain technology ensuring secure carbon credit transactions. Challenges in managing big data include volume, security, and the lack of global standards, with solutions involving infrastructure improvement, international collaboration, and AI-driven analysis. Key takeaways emphasize the importance of a statistical mindset, promoting data literacy from early education, adapting curricula, establishing a robust data infrastructure, ensuring data accuracy and security, and recognizing blockchain's role in transparent data transactions.