Yulia Gel | Statistics and Actuarial Science

Associate Professor

Department of Statistics and Actuarial Science
University of Waterloo
Waterloo, Ontario
CANADA N2L 3G1

Phone: (519)888 4567 ext.36190
FAX: (519) 746-1875

E-mail:
Office: Math & Computer 6104B

Research and Scholarly Activity

Professor Gel's research interests are quite broad and include but are not limited to time series analysis, spatio-temporal modelling, and more recently non-parametric statistics. The main applications include weather forecasting, analysis of fMRI data, and analysis of data coming from various law cases, for example in the areas of security law and equal employment.

Her main interest in time series is estimation and inference for infinite-dimensional or "long'' autoregressive models. The problem is of practical importance because the class of such models includes (but is not limited to) all causal invertible autoregressive moving-average (ARMA) models. "Long'' autoregressive models may be used as an alternative to models proposed by information criteria (AIC, BIC, MDL etc.) in on-line parameter estimation problems, i.e., when the length of the observed data sample is not known a priori and may increase indefinitely; and as an approximation to "long'' memory models. Such an approach can be considered as a smoothed version of the currently utilized parameter estimation and model selection procedures.

Professor Gel's research in spatio-temporal modelling addresses the application of existing and the development of new statistical methods for improving the
prediction quality of weather forecasts that are primarily intended for local users, e.g., farmers, transportation agencies, energy companies, schools, and the general public. This project is highly application-driven and involves various areas of statistics (time series analysis, spatial statistics, Bayesian inference etc.). The potential application of the methodology goes far beyond the modelling of meteorological events and might include the spatio-temporal analysis of functional magnetic resonance data, or geophysical and oceanographic data. Currently the two main targets are the modelling of bias in temperature and pressure weather forecasts and a method of statistical ensembles for mesoscale weather prediction.

Recent Publications

Gel, Y. (2009) Test of fit for a Laplace distribution against heavier tailed alternatives. To appear in Computational Statistics and Data Analysis. 11 pages.
Gastwirth, J.L., Gel, Y. and W. Miao (2009). The Impact of Levene's Test of Equality of Variances on Statistical Theory and Practice. To appear in Statistical Science. 22 pages.
Gel Y., W. Miao, J. L.~Gastwirth (2009). The effect of dependence between observations on the proper interpretation of statistical evidence. Law, Probability and Risk, 8(1), 25--38. (more)

Biography

Recently, Professor Gel was employed as a statistical consultant on time series data analysis for a security law case and got involved in assessing the potential effect of deviations from the assumptions underlying various statistical tests. This eventually led to a number of projects on developing new non-parametric and goodness-of-fit tests, e.g., directional normality tests against heavy-tailed alternatives, a robust test of symmetry about the unknown median, and a two-stage procedure for the Wilcoxon and Shapiro-Wilk tests when the assumption of independence among data is violated.

After graduating in 2000, Professor Gel had two postdoctoral appointments in the University of Washington and in Malardalen University in Sweden, followed by
a position as a visiting assistant professor in George Washington University. She joined the University of Waterloo in 2004. From 2003 to 2005 Professor Gel was
a statistical consultant for the law firm Watchtell, Lipton, Rosen, and Katz, in New York City.

Currently Professor Gel primarily collaborates with the National Weather Service in Washington, DC (statistical methods in mesoscale weather forecasting), the Department of Radiology at the University of Washington (spatio-temporal modelling of fMRI data), and George Washington University (legal statistics and non-parametric methods). (top)