Spatial prediction filters as dip filters

A spatial prediction filter can take the form of a dip filter. Using linear least squares we can estimate the coefficients (a,b,c,d,e) in the 2D filter

		 		a
		 		b
		1 		c
		 		d
		 		e

Fitting the filter to two neighboring traces that are identical but for a time shift, the filter (a,b,c,d,e) should turn out to be the negative of an interpolation filter. Ideally you might see (-1,0,0,0,0) or (0,0,-.5,-.5, 0). But if the two channels are not fully coherent you expect to see something like (-.9,0,0,0,0) or (0,0,-.4,-.4,0).