Statistics Colloquium
November 15 (Fri) 11 a.m.
Alavi Commons Room, 6625 Everett Tower
Estimation of smoothness of a stationary Gaussian random field
Dr. Chae Young Lim
Department of Statistics
Michigan State University
Analysis on smootheness (or roughness) of time series and spatial data has a wide application in various disciplines. Fractal or Hausdorff dimension is a measure of smoothness and several fractal-based methods are available. For a stationary Gaussian random field, the decay rate of the spectral density as frequency increases is related to the fractal dimension. We propose an estimator of the decay rate using periodogram when the observations are on a grid and investigate its theoretical properties under infill asymptotics. From the property of the estimator we found, we also propose an algorithm to reduce bias. We compare our approach with other existing methods via simulation. Real data example on arctic sea-ice profiles is also presented.
All statistics students are expected to attend.
.
Past colloquiums
