Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation

>> One thing you might want to pursue that would interest me is "what kinds of data lend themselves to interpolation, regression, or any other kind of smoothing, and what kinds do not." <<

Let me do a cut*paste here for a book that gives you the mathematical criteria for continuous functions.

Interpolation: Second Edition by J. F. Steffensen  Price $14.95 ISBN10: 0-486-45009-0

In the mathematical subfield of numerical analysis, interpolation is a procedure that assists in "reading between the lines" in a set of tables by constructing new data points from existing points. This rigorous presentation includes such topics as displacement symbols and differences, divided differences, formulas of interpolation, much more. 1950 edition.

Statistics has a lot of techniques for missing data when it is measured on a nominal, ordinal or rank scale. They can be okay, but when they stink they really stink. I cannot think of a good book on missing data! Might be a good topic for my next book.

(I am doing this from memory) Years ago, AMERICAN STATISTICIAN (?) had an article on the huge "Black Hispanic" population in Los Angeles according to the 1990 US Census. Hispanics (read: illegal Mexicans) had displaced upwardly mobile Blacks in many census tracts. Illegals do not like to fill out forms, do not speak much English or avoid all contact with government agents, so you got a lot of missing data. The missing ethnic data was filled in by looking at prior data and data in adjacent tracts. Ergo, they are Black!

The error was discovered by someone at UCLA who looked at the public schools, emergency room stats and common sense.

But statistics is the place to start for a theory of missing data. I would also look at resampling techniques, too. Received on Thu May 03 2007 - 18:23:43 CEST