The frequency-domain properties of many biomedical time series contain valuable information. These properties are characterized through its power s pectrum, which describes the contribution to the variability of a time series from waveforms oscillating at different frequencies. Practitioners seeking low dimensional summarymeasures of the power spectrum from a population often partition frequencies into bands and create collapsed measures of power within these bands. However, standard frequency bands have largely been developed through subjective inspection of time series data and may not provide adequate summary measures of the power spectrum for a given population of interest. This proposal seeks to establish a new framework for adaptive frequency band estimation and analysis for replicated time series, thus bridging an important gap between the analysis of spectral information from a single time series and the analysis of spectral information within a population. The four specific aims associated with the effort are: (1) to develop a frequency band estimation method for replicated, stationary signals that best preserves variability across replicates within a population, (2) to develop a local frequency band estimation method for replicated, nonstationary signals that best preserves time and replicate-varying behavior within a population, (3) to develop a frequency band estimation method for replicated, multivariate signals that best preserves the characteristics and interrelationships between individual components and (4) to develop a suite of user-friendly analytical tools across multiple software platforms. Monte Carlo simulation studies will be conducted to explore the empirical prope rties of the proposed methods and to compare their performances to the use of traditional frequency bands. The investigators will use these new methods to analyze a range ofbiological signals, including heart rate variability, pupil dilation, and MRI, from three existing studies to address a variety of biological and clinical questions. The impact in practical investigations is expected to be substantial, equipping practitioners with justified optimal tools for analyzing data collected from a broad spectrum of scientific and biomedical studies.
This proposal will design practical statistical procedures for identifying frequency band summary measures of biomedical time series data that optimally characterize oscillatory patterns for a population of interest. The investigators will use these new methods to analyze biological signals from three existing studies and provide practitioners with optimal tools for analyzing data from a broad spectrum of biomedical studies.
