I have a sparse, rectangular array where sparse means that missing entries in the table are non-existant, not zero. The array is "patterned" such that on any one line or column, there is one and only one short segment of existing data.
One dimensional curve fitting of an individual line or column demonstrates that each such segment is well represented by a cubic polynomial over its individually limited range.
Is there any means by which I can reduce this array of numbers (200x740) to a managable analytical function? Two hundred cubics (4 parameters plus two limits) would do it, but I need something more compact, if possible. I haven't any idea of how to procede, so all suggestions are very welcome.