140 likes | 413 Views
Fin250f: Lecture 9 Spring 2010 Brooks, chapter 12.1-12.4. Monte-carlo and Bootstrapping. Outline. Computer simulation Monte-carlo Bootstrapping. Why Simulations?. Small samples Complex expressions (no analytics) Ease of use (hard to get analytics). Examples from Finance.
E N D
Fin250f: Lecture 9 Spring 2010 Brooks, chapter 12.1-12.4 Monte-carlo and Bootstrapping
Outline • Computer simulation • Monte-carlo • Bootstrapping
Why Simulations? • Small samples • Complex expressions (no analytics) • Ease of use (hard to get analytics)
Examples from Finance • Risk management • Value-at-Risk/Stress testing • Option pricing • Trading rules and data snooping • Stationarity testing
Monte-carlo procedure • Postulate "data generating process" DGP • Simulate with random number generator y = b*x + e • Estimate b • Store distribution • Examples: simplemc.m, snooprule.m
Random Number Generators • Fake random numbers • Random seed to start • Matlab and seeds • newSeed.m
Bootstrapping • Use actual data • Draw new set of data independently • Stock returns • Write on paper • Throw in urn • Draw returns at random with replacement • Sample.m • genRandomWalk.m
Bootstrapping a Regression • Two methods • Method 1 • Resample (y,x) pairs • Method 2 • Resample e = y-bx residuals • Then regenerate y from scrambled e
Power of the Bootstrap • No distributional assumptions • Data speaks for itself
Problems with Bootstrap • Outliers in the data • Too little data • Nonstationary data • Nonindependent data • Max's and mins
Antithetic Sampling • If distribution is symmetric about zero • Simulate with bootstrap • Randomly flip sign • Like having twice as much data
Bootstrap Examples • bootcapm.m • bootquantile.m
Monte-carlo Examples • mclinearsnoop.m • In sample/out sample • snooprule.m • Trading rule snooping • Best rules out of a group • mcmse.m