Gary W. Oehlert

I enjoy applied statistics. I enjoy being part of projects which interpret data to answer important questions. Some of the questions on which I have worked include: How do we interpret and improve tests for mutagenic compounds? Is the ozone layer being depleted? Is acid rain on the increase? Is the greenhouse effect already at work? How accurately can we now estimate the historic acidity of lakes? Which parts of the brain are active during certain tasks? Real applied problems rarely admit standard statistical solutions. There is always something unusual or unique. Each problem is a new challenge with a new blend of theory and methodology (both old and new), and new computing techniques to implement the methodology may be required. Consider, for example, the problem of detecting trends in temperature. Data are available for nine atmospheric levels for various time periods at over 500 stations unevenly spread around the world, and most stations have incomplete data. One theoretical question is, "How do you model such data?" Clearly, data values close in time or space are correlated, and this needs to be accommodated. Fitting the model to the data is a methodological problem. What technique should be used? What attributes--standard errors, for example--will the estimates have? Can we assess how well the model fits? How can we summarize the results in a pleasing graphical fashion? Finally, the computer must be programmed to efficiently process a mountain of data using a new model and fitting technique. Applied statisticians get to learn enough about a subject area to understand what is going on with the science. But we also learn to abstract the questions and situations, so that we can see that changes in the ozone layer and changes in brain activation actually have a lot in common. The joy of applied statistics is that it allows you to combine many aspects of statistics with a field such as climatology or medicine. And that combination can't be beaten.