Related Experiment Videos
Exploring functional data analysis and wavelet principal component analysis on ecstasy (MDMA) wastewater data
Stefania Salvatore1, Jørgen G Bramness2, Jo Røislien2,3
1Norwegian Centre for Addiction Research, University of Oslo, Oslo, Norway. stefania.salvatore@medisin.uio.no.
BMC Medical Research Methodology
|July 14, 2016
Summary
Functional principal component analysis (FPCA) offers a robust method for analyzing wastewater-based epidemiology (WBE) drug use data. FPCA, particularly with Fourier basis functions, proves stable and accurate for tracking community drug trends.
Area of Science:
- Environmental Science
- Epidemiology
- Data Analysis
Background:
- Wastewater-based epidemiology (WBE) monitors community drug use.
- Novel analytical methods are needed for WBE data.
- Functional principal component analysis (FPCA) is explored for WBE.
Purpose of the Study:
- To investigate FPCA for analyzing temporal WBE data.
- To compare FPCA with traditional principal component analysis (PCA) and wavelet principal component analysis (WPCA).
- To assess the stability and sensitivity of FPCA to missing data.
Main Methods:
- Analyzed temporal wastewater data from 42 European cities.
- Extracted temporal features of ecstasy (MDMA) using FPCA, PCA, and WPCA.
- Explored FPCA stability via bootstrapping and sensitivity analysis.
Main Results:
- The first three components explained 87.5-99.6% of temporal variation.
- Extracted temporal features were consistent across PCA, FPCA, and WPCA.
- FPCA with Fourier basis and common-optimal smoothing showed highest stability and least sensitivity to missing data.
Conclusions:
- FPCA is a flexible, robust method for analyzing temporal WBE data.
- FPCA is superior to WPCA in capturing rapid temporal changes.
- FPCA with Fourier basis and common-optimal smoothing is recommended for WBE data analysis.