Jeremy Schiff: Numerical Methods for Mathematical Finance (88-636)

Comments on grading of the computer test

Each question was graded out of 50. Within each question I gave points as follows:

25 points for a basic functioning program that does the minimum needed (compute the value of the option for a single value of c in question 1, compute the value of the contract in question 2, solve the PDE in some manner in question 3).
10 points for relating to the issues in writing the program. For example: choice of method where more than one was available, choice of numeric parameters ("N" and "M" in questions 1 and 2, "h" and "k" in question 3; also choice of a sensible number of values of c in question 1), deciding what results are worth looking at.
15 points for giving results and discussion. In both questions 1 and 2 the results raise questions that need to be answered. For example in question 1, many people found results for different c's without using common random variables, and thus their graphs went up and down; but very few asked whether this is a real phenomenon (no) or caused by numerical error (yes). In question 2, more people did relate to the question of why the value of the contract goes up when the sigma's are changed as suggested, but very few investigated it in any detail (use the computer!).