Psychology
414 & 505
Monte Carlo Techniques: Bootstrapping, Randomization and Experimental Statistics
This course is designed to teach computer simulation (Monte Carlo)
based statistical techniques, with special emphasis on practical
solutions to problems faced by behavioural scientists attempting
to publish analyses of small, noisy, data sets in peer reviewed
journals.
Three specific topics will be covered:
- Randomization: A method for hypothesis testing that is
remarkably assumption-free, robust and powerful. Understanding how
randomization testing works can let you construct novel tests when
you find yourself with data that doesn't seem testable using
classical statistics.
- Bootstrapping: A method which allows robust confidence
interval estimation. Intelligent use of the bootstrap may allow you
to convince journal editors, and anonymous referees that your
non-significant result is not a reflection of poor data collection,
or inadequate sample sizes, but a real, interesting and important
publishable result.
- Experimental statistics: Computer-intensive simulations
of statistical tests can be used to explore how robust, or fragile,
those tests are to violations of their assumptions, or empirically
investigate other non-obvious properties of tests.
Students will perform a modest amount of computer programming, but
no previous experience in programming is required. Software will be
provided.
Log onto E-Learning
to download datasets, lecture notes, class readings etc.
Back to Prof. Hurd's
homepage