McMASTER UNIVERSITY STATISTICS SEMINAR

Week of November 16 - 20, 1998

SPEAKER:	Dr. Hugh Chipman Department of Statistics and Actuarial Science, University of Waterloo
TITLE:	"Tree Models: Roots and Recent Branches"
DAY:	Wednesday, November 18, 1998
TIME:	3:30 p.m. [Coffee in BSB-202 at 3:00 p.m.]
PLACE:	BSB-108

SUMMARY

Trees provide flexible and often interpretable ways to model data. By using one or more explanatory variables and a tree-structured set of questions, tree models divide a population into similar groups. This talk will provide an introduction and overview of tree models, including CART (Breiman et. al., 1984) and C4.5 (Quinlan, 1993). Topics will include tree construction and validation, and some more recent methods for the identification and selection of trees when many different trees may fit the data well.

ABOUT THE SPEAKER

Dr Hugh Chipman received his Bachelor's degree from Acadia University and his Master's and PhD from the University of Waterloo under Jeff Wu and Michael Hamada. His thesis Bayesian regression methods for ordered categorical data won the Pierre Robillard Award of the Statistical Society of Canada for 1994. He was a faculty member at the University of Chicago Graduate School of Business from 1994-97, and has been at the University of Waterloo since that time. His research interests include tree models, model selection in regression problems, and data mining.

REFERENCES

The following references have been provided by Dr. Chipman to be used as background for his talk. References [1] & [2] are on reserve at Thode Library (STATS 770: Statistics Seminar).

[1] Clark, L.A. & Pregibon, D. (1992) "Tree-Based Models", in STATISTICAL MODELS IN S, J. M. Chambers and T. J. Hastie, Eds, Wadsworth & Brooks/Cole

[2] Chipman, H. A., George, E. I. & McCulloch, R. E. (1998) "Bayesian CART Model Search (with discussion)", JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION 93, 935-960.

[3] Chipman, H., George, E. & McCulloch, R. (1998) "Extracting Representative Tree Models From a Forest", Working Paper 98-07, Department of Statistics and Actuarial Science, University of Waterloo. (available online at http://setosa.uwaterloo.ca/Stats_Dept/techreports/node5.html)