Yale ECON 561-417 - Computational Methods in Economics: General Points

Unformatted text preview:

Econ 561a (Part I)Yale UniversityFall 2005Prof. Tony SmithComputational Methods in Economics:General Points• Three modes of science: theoretical, computational, empirical.Computational work shares elements of both theoretical and em-pirical work.• Two meanings of computational economics: computation as atool for doing standard economic theory vs. computation as amodel for how economic actors behave.• Fundamental shortcoming of the first kind of computational eco-nomics: lack of error bounds. How far apart are the computa-tional model and the theoretical model?• Computation produces “data” which can be used to generatetheoretical conjectures. These conjectures could be exact (theo-rems) or approximate (systematic patterns).• Use computation to measure quantitative magnitudes (e.g., therelative sizes of opposing effects).• Hamming’s motto: The Goal of Computing is Insight, Not Num-bers.1Econ 561a (Part I)Yale UniversityFall 2005Prof. Tony SmithRules of Thumb for Doing GoodComputational Work in Economics1. Start with the simplest possible model, preferably one with ananalytical solution.2. Add features incrementally.3. Never add another feature until you are confident of your currentresults.4. Use the simplest possible methods.5. Accuracy is more important than speed or elegance.6. Use methods that are as transparent as possible (i.e., methodsfor which the computer code reflects as closely as possible theeconomic structure of the problem).7. When you learn (or de velop) a new method, test it on the sim-plest possible problem, preferably one with an analytical solution.8. Dan Bernhardt’s rule: If you have n errors in a piece of code andyou remove one, you still have n errors. Scrutinize your r esults,even if they look right. Look for anomalies. Assume your codeis wrong until proven otherwise.9. Graph, graph, graph. Two-dimensional graphs are more infor-mative than three -dimensional graphs.210. Be able to replicate all of your intermediate and final results in-stantly. Save exact copies of the code used for each run, togetherwith inputs and outputs.11. Watch the computations as they proceed.12. Exploit homotopy.13. Learn a fast language, such as C or Fortran.14. Look for hidden structure. Always compute (and print out) afew more numbers than you need.15. Get good initial conditions.16. Use one-dimensional algorithms as much as possible.17. Use algorithms that can be “tightened”.18. Avoid black boxes. Understanding how the algorithm works iscritical to interpre ting the results.19. Remember that programming is a creative activity analogous towriting a sonnet or composing a sonata (a static, visual represen-tation of a process). Craft your programs. Strive for efficiencyand elegance in your computer code. Develop a style. Practicestructured programming, i.e., write code that reflects the struc-ture of the algorithm.20. Don’t program when you are tired. Don’t program too


View Full Document

Yale ECON 561-417 - Computational Methods in Economics: General Points

Download Computational Methods in Economics: General Points
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Computational Methods in Economics: General Points and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Computational Methods in Economics: General Points 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?