Abstract:
|
Climate change poses significant challenges to the soil ecosystem, with profound implications for many aspects of life. In addition, much of the data collected in the field pose interesting statistical challenges due to the presence of temporal dependence. Due to inherent dependencies in time series, many of the classical statistical inferential methods are inadequate. To overcome this difficulty, a functional anova approach may be utilized to compare groups of time series. We present an application in which climate change was experimentally simulated in the montane meadows, resulting in time series measurements of soil moisture and temperature. After smoothing the discrete measurements to obtain functional curves, an anova test for functional data is performed through a parametric bootstrap procedure to test equality of mean curves between treatment groups. Extending this procedure, we develop a method to test for interaction between treatments. We also develop and illustrate novel visualizations of the tests, which not only provide another facet in understanding the significance of the tests, but also allow for identification of when significant differences occur over time.
|