Associate Professor David ScottSports PredictionIf you are interested in my Super Rugby and NRL predictions they can be found on the Statistics Department blog Stats Chat. The method used for these predictions is very simple. To predict the outcome of a game I estimate the margin as the home team rating minus the away team rating, plus the home ground advantage value. If the margin is positive, the home team is predicted as the winner. If it is negative, the away team is predicted as the winner. Ratings for the teams are adjusted after the game using the error correction form of exponential smoothing. Obviously there is some initial estimation of team ratings, and some choice of the optimum parameter values for the exponential smoothing constant and the home ground advantage. Once these tasks are complete, the prediction routine can be run automatically. It's not rocket science, just statistics. TeachingThis year I am teaching STATS 779 Professional Skills for Statisticians jointly with James Curran, STATS 320 Applied Stochastic Modelling, and part of STATS 370 Financial Mathematics. STATS 779 ranges broadly over a range of software which is of use to professional and research statisticians. STATS 782 is a predecessor of STATS 799. For both courses I have collected a range of documentation which I have found useful both in my own work and in teaching. Go to the pages STATS 779 or STATS 782 to access this material.
Software: Reproducible ReasearchMy most recent package is hwriterPlus which is a package for Reproducible Research or automated report writing. It extends the package hwriter which is on CRAN. Here is the description for hwriterPlus. "This package extends the package hwriter providing facilities such as the inclusion of output from R, the results of an R session, the display of mathematical expressions using LaTeX notation and the incorporation of SVG graphical objects." hwriterPlus contains an extensive example showing how it may be
used. The file used is called There is also a vignette describing the package, named
Software: DistributionsI have developed a number of packages for distributions which are available on CRAN. Currently these are DistributionUtils, GeneralizedHyperbolic, SkewHyperbolic, and VarianceGamma. My R package HyperbolicDist is still available on CRAN, but deprecated since it has been replaced by DistributionUtils and GeneralizedHyperbolic. Development versions of my packages are all available on RForge as part of the Rmetrics project. This link Rmetrics packages on RForge will take you to those packages. In addition to the packages mentioned above there is a package for the normal Laplace distribution on RForge which is not yet on CRAN. There are some other functions which I have not yet included in my packages. Most of them are alternative methods of simulation of generalized hyperbolic observations. These functions may be downloaded and used provided the the source is acknowledged.
OtherI use BEAMER for the production of pdf slides using LaTeX. Here are some example slides and the LaTeX files used to produce them. These were used in a presentation to staff and students in the Department of Statistics on June 10, 2008.
