Department of Statistics, The Ohio State University
Statistics and Biostatistics Colloquium Series

We collaborate on several microarray experiments, with the first task being to design these experiments. At this point we have extended and generalized the results of Kerr and Churchill (2001) on design optimality within Balanced Incomplete Block and Row-Orthogonal designs for larger numbers of varieties (samples) and single factor experiments. Future designs needs are for multifactorial experiments. We are beginning to consider row-column designs for blocking in two directions (arrays and dyes). We also consider other design issues (e.g., balance between factors, randomization). Researchers are beginning to use two-step mixed model approaches for the identification of factors influencing gene expression. We are investigating under what conditions the two-step and one-step analyses yield the same test statistics, where two-step analysis is only approximate and to what extent, and how to implement one-step analysis more efficiently. We are also working on sample classification methods, which are suitable for training data with sample misclassification, in particular modified maximum likelihood logistic regression and Bayesian binary regression. Finally, we are interested in regulatory and metabolic pathway inference. Here, we are investigating factor analysis as a tool for metabolic pathway identification, and a combination of 'Genetical Genomics' and Bayesian networks for learning regulatory pathway structures.

Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.