Random lines in a plane

UTILITY OF THIS RESULT

At present we are accustomed to estimating statistical properties of seismic data by computing a 2-D prediction-error filter. This filter is needed to interpolate and extrapolate missing values.

Knowing that the prior spectral estimate is not a constant but instead is $1/k_r$ suggests a procedure that is more efficient statistically: By more efficient, I mean that a simpler model should fit the data, a model with fewer adjustable parameters.