| Keywords [eng] |
COVID19, Media Sentiment, Functional Data Analysis, Functional Canonical Correlation Analysis, FunctiononFunction Regression COVID19, žiniasklaidos nuotaikos, funkcinė duomenų analizė, funkcinė kanoninė koreliacijos analizė, funkcinė regresija funkcijai pagal funkciją |
| Abstract [eng] |
This thesis applies Functional Data Analysis (FDA) to model and interpret the weekly spread of COVID-19 in Lithuanian municipalities from 2020 to 2022, treating all relevant processes as smooth functional trajectories. To better understand the influences behind COVID-19 dynamics over time, the analysis examines the roles of weather conditions (temperature and absolute humidity), vaccination rollout, and Lithuanian media sentiment. The methodological framework integrates Functional Principal Component Analysis (FPCA), Functional Canonical Correlation Analysis (FCCA), lagged FCCA, and function-on-function regression into a single workflow, representing a relatively rare application of a unified FDA approach to epidemic processes in Lithuania. The results show that seasonal meteorological patterns were the dominant structural drivers of COVID-19 cases, that vaccination effects primarily reflected national synchronisation rather than municipality-level variation, and that media sentiment offered a modest but anticipatory signal of rising case numbers. These findings highlight how FDA can reveal coherent temporal relationships across epidemiological, environmental, and informational processes, and demonstrate its potential for improving epidemic monitoring in settings with rich time-series data. |
