CMStatistics 2022: Start Registration
View Submission - CMStatistics
B1646
Title: Functional factor model for density functions Authors:  Israel Martinez-Hernandez - Lancaster University (United Kingdom) [presenting]
Abstract: Air pollutants are the most studied phenomena due to their impact on our daily life. In particular, the study and understanding of different sources of particulate matter (PM). Recent evidence has shown that a mixture of particles from different sources can have different toxicity and health effects. For this reason, particle number size distribution (PNSD) measurements have received much attention and are used to investigate PM sources. Due to the high correlation and a large amount of data, current models are highly computationally demanding or use a simple surrogate model. We propose to use a functional data analysis approach. PNSD can be naturally considered as a sequence of density functions over time. With this motivation, we propose a functional factor model that considers the time dependency and overcomes several challenges the current models face. We will illustrate our methodology using apportion PNSD measured near London Gatwick Airport (UK). Our model can identify the most common PM sources and provides accurate information on how each source contributes to the total PM.