Pierre R.l. Dutilleul

Biography

Pierre Dutilleul’s background is in mathematics and statistics, and his research interests are in statistical inference (estimation and testing) and applied statistics, the domains of application including plant science, ecology and the environmental sciences, agronomy and crop science, forestry and dendrochronology, soil science and seismology. Accordingly, he is Professor in the Department of Plant Science and Associate Member of the Department of Mathematics and Statistics and of the McGill School of Environment. In Google Scholar, Dr. Dutilleul is most known (600+ citations) for his modified t-test for correlation analysis with spatial data. Professor Dutilleul is also known for his innovative phytometric research work, in which his group is making use of a computed tomography (CT) scanner to collect 3-D spatial data on plant structures and analyzing them statistically (see, e.g., his interview to the Science magazine in February 2006 and the Radio-Canada Découverte reportage made in February 2007); this work has since been extended to the studies of soil and wood structures. Pierre Dutilleul has authored ~150 peer-reviewed publications and one book (“Spatio-Temporal Heterogeneity”, Cambridge University Press, 2011) and has coordinated from beginning to end the e-book project “Branching and Rooting Out with a CT Scanner” (Nature Publishing Group/Macmillan, 2016).

Research Interest

In statistical sense and simply put, heterogeneity may concern the mean or variance parameter of the distribution of a random variable, or be related to the autocorrelation function of a stochastic process. When the value of the mean or the variance is susceptible to change, or variability is measured from observations that are partially dependent on each other because they are autocorrelated, in time or space, there is potential for a heterogeneity analysis, starting with the experimental design (Dutilleul, 1993a, Ecology). This opens the door to a lot of interesting situations and problems! A modified t-test (Dutilleul, 1993b, Biometrics) provides a solution to the problem of assessing validly the correlation between two autocorrelated spatial processes, and was followed by a modified F-test and other modified t-tests in the contexts of multivariate and multi-scale analyses