Both Case Studies for the SSC 2004 Annual Meeting are now posted.
|
Statistical computing, statistical software packages; working with large data sets; exploratory data analysis; graphical methods; statistical consulting practice.
Monday 11:30-13:20 in BSB-103, Wednesday 11:30-12:20 in BSB-106
Monday 13:30, Wednesday 10:30, other times by appointment.
All assignments must be submitted no later than noon on Wednesday 28 April 2003, the final day of the undergraduate examination period.
Consider membership in herd 1 as a binary y-variable. Find and fit tree and logistic regression models. Split the data into training and validation samples, possibly keeping the numbers from each herd equal in both samples. Consider the complexity of the model, making sure you are not over-fitting. Give box plots, ROC curves and other graphical summaries. Do your work in Splus or R, and in SAS Enterprise Miner. Write a report explaining your results.
Consider the data on tree diameters handed out in class on 2004-03-24. Does a three-parameter Weibull distribution with location parameter = 2.5 fit the data adequately? If not, does a mixture of two three-parameter Weibull distributions, each with location parameter = 2.5, fit? Remember to adjust for the differential sampling of small trees. Prepare a consulting report to explain your methods and your findings to the client. (Read Chatfield, Part I, Chapters 10 & 11 for suggestions on report writing.)