2:00-5:00 Pre-Conference Short Course Tutorials
(separate registration required)
Using Classical Statistics and Experiment Designs, Reduce Noise in Your Next Microarray Experiment
Mr. Thomas J. Downey, President, Partek, Inc.
Microarray data contains treatment and/or phenotype effects embedded in a sea of technical and biological noise. This workshop will demonstrate how to use proven statistical methods of experiment design and data analysis to reliably identify biological effects of interest while controlling and removing noise due to biological and technical nuisance effects. Attendees will learn how to conduct completely randomized block designs to isolate and remove batch effects due to dye, reagent lots, chip lots, and hybridization batches, clearly revealing the signals from the biological factors of interest. In addition to calculating correct p-values, statistically correct estimates of ratios and fold-changes will be examined from a classical statistical perspective. Examples will be included using published two-color cDNA and Affymetrix experiments.
Who Should Attend & Why:
The workshop will be most useful for scientists who wish to enhance their understanding of classical statistical methods of experiment design and analysis of variance (ANOVA) and serves as an excellent "warm-up" for the remainder of the conference.
What Attendees Will Learn:
Attendees will learn how to use fundamental concepts of experimental design and analysis of variance in addition to more advanced techniques such as randomized block designs, mixed model analysis of variance, and nested cross-validation for accuracy estimation of predictive models.
Taking Multivariate Statistical Methods out of the Black Box
Ms. Janis Dugle, Senior Research Scientist (Statistics), Ross Products Division, Abbott Laboratories
The multivariate statistical methods of Principal Components Analysis (PCA), Partial Least Squares (PLS) and Discriminant Analysis (DA) have become popular tools of the burgeoning biosytems and "omics" sciences (genomics, proteomics, metabonomics, etc.). However, they often appear as "black box" techniques that leave the scientist in the dark about what has been done to his/her data. This mini-course provides a non-technical introduction to these methods. Many examples of PC, PLS and DA are given (sometimes on the same data), which allows comparisons of their capabilities, limitations, and commonalities. Many sources of data are used to illustrate the wide range of application of these techniques, and the resulting graphical output gives a hands-on feel to large data sets. The course ends with a grand finale of a marker search through data with thousands of measurements on 116 samples. ! This class is enjoyable and easily understood, regardless of your statistical background.
Who should attend?
This course is designed for anyone who wants a non-technical, understandable, highly visual overview of PC, PLS, and DA. It is appropriate for the researcher who is interested in seeing what statistics can and can't do, but who is not concerned with technical statistical derivations.
Do you want to be able to "talk the talk" in multivariate methods? This is the place. Find out what they are doing with your data, and what other options may be available