From jpretti@barrow.uwaterloo.ca Thu Mar 12 09:12:32 1998 Date: Tue, 10 Mar 1998 14:47:18 -0500 (EST) From: John-Paul Pretti To: hwolkowi@orion.math.uwaterloo.ca My comments to the students: The second case study was done better than the first. Most groups received a better grade on the "university" study. The following is an approximate breakdown of the marks: 20 Communication - summary - conclusions and recommendations (often omitted) - organization - usability (best treated as a report to your boss) 20 Software - GAMS implementation and documentation - don't forget to use meaningful variable names - explain how your test data was generated - is your solution valid ? how can you test this ? if the problem is large ? 20 Sensitivity Analysis - simply defining shadow prices is not enough - give concrete examples specific to your study - highlight the most important implications - give recommendations - even if you can't answer a question, do pose it !! 20 Model - list all assumptions (I'm sure you all made at least 10) - discuss your variables, do you force them to be binary ? 10 Size of Problem - try a full scale problem (size 100,500) - when/does it blow up ? - discuss the trade-off between accuracy and tractability - what does the objective fn really mean ? 10 Extensions - think of as many as you can - try implementing at least one The best studies were brief and clear. An important skill to learn is to determine what things are most important and highlight them. Simply putting everything you think of on paper is not the best approach. Good luck with the project ! John-Paul Pretti (J.P.) University of Waterloo Department of Combinatorics & Optimization MC 5136A (519)888-4567 x6895