All Times ET
Virtual
Spatio-Temporal Analysis of Particulate Matter Based on Quantile Factor Model (304211)
Joonpyo Kim, Seoul National University*MINJI KIM, Seoul National University
Hee-Seok Oh, Seoul National University
Keywords: Spatio-temporal data analysis, quantile analysis, approximate factor model
The ultimate goal of this study is to analyze the hidden factor structure of large-scale fine particulate matter (PM 2.5) data observed hourly for five years at 103 stations around Korea. To achieve this goal, a novel quantile factor model (QFM) is carried out for spatio-temporal analysis of PM 2.5 values.
Our analysis assumes factors to have dependence, expecting that the factor structure efficiently summarizes the vast amount of serial correlation within and across individual processes. Also, our quantile based approach enables us to detect heterogeneous effects of factors at different quantiles of the data and not be restricted only on the average effect. We further expand the quantile analysis to some extremal levels that capture the tail variables of the data based on extreme value theory.
In sum, the main purpose of this study is to apply the QFM to analyze the joint tail behavior of PM 2.5 anomalies.