Rafe Donahue's Home Page

My cv (pdf approx. 83 k, last updated 02 April 2008)

A link to my page at Vanderbilt Unviersity Medical Center

My mathematical genealogy (pdf approx. 87 k, as of 2005.10.22) Data used to make this graphic were pulled from The Mathematical Genealogy Project, in particular, starting from my page in the great tree.

Bivariate Baseball Score Plots: This interactive website was launched at the 2007 Joint Statistical Meetings in Salt Lake City in late July 2007. The handout (pdf approx. 147 k) from that presentation and a blog for adding insightful commentary are also available. (Added 2007.08.04)

When Summary Measures Aren't Enough: A demonstration of short-comings of simple summary measures (Added 2007.01.24)

Parental Concerns Questionnaire in Autistic and Typical Children: Descriptor page and link to pdf

Standardizing Clinical Study Designs For Accelerating Drug Development --- from the Drug Information Journal , Volume 31, 1997: Complete article (pdf approx. 40 k)

Some Commentary On Maximizing The Statistician’s Value To The Pharmaceutical Industry --- from the Drug Information Journal , Volume 34, 2000: Complete article (pdf approx. 40 k)

The Development of a Darwinian Scheme for Estimating Rankings Under a Stochastic Ordering --- a paper I presented at the American Statistical Association Annual Meeting in Indianapolis, August 2000: Complete article (pdf approx. 145 k)

How Statisticians Think and Why It Matters --- a paper I presented at the Drug Information Association Annual Meeting in Denver, July 2001: Complete article (pdf approx. 100 k)

Comparing R&D and Commercial: Are All Data Created Equal? --- a paper I presented at the Midwest Biopharmaceutical Statistics Workshop in Muncie, Indiana in May 2002: Complete article (pdf approx. 82 k)

Statistics in Clinical Research: Godsend or Necessary Evil? --- a paper I presented at the Regulatory Affairs Professionals Society Clinical Issues Meeting in San Diego in April 2003: Complete article (pdf approx. 96 k)

Data Issues in Multinational Trials --- another paper I presented at the Regulatory Affairs Professionals Society Clinical Issues Meeting in San Diego in April 2003: Complete article (pdf approx. 83 k)

An Overview of Data Mining in Drug Development and Marketing --- a paper I presented at the Midwest Biopharmaceutical Statistics Workshop in Muncie, Indiana in May 2003: Complete article (pdf approx. 66 k)

A cute little application of a sparkline -like graphic in text (pdf approx. 7 k) . The plots were made with SAS/Graph and the text was done in MS Word and then run into pdf. The data came originally from my son's 5th grade math book and a question regarding using the mean or median to represent the data. I said that neither one is optimal; what matters is seeing the distribution. So, I put the data into a little graphic and made up a little story.

An actual use of PowerPoint. When I was between gigs, I no longer had access to my favorite graphics production facility, SAS/Graph. In a pinch, I used Excel to draw the plot and then used PowerPoint to make an annotated story line. The result is a chronicle of weight readings for nearly a year (ppt approx. 70 k). (The file is a PowerPoint file, so you need PowerPoint, or the PowerPoint viewer, to view it. The viewer is available via a free download at Microsoft's home page. Once there, do a search on 'powerpoint viewer' and follow the instructions.) One slide; print it out, 8.5x11, landscape. The graphic is not quite what I would have liked, as my Excel skills are somewhat lacking. I would have liked to be able to shade the individual months to set them apart from each other, but this will have to do. Note that, with some effort, we can get decent displays out of PowerPoint and Excel. The hands-on manipulation of the elements, however, drives me batty. I dread collecting more data for fear of having to eventually redo all my work!

Graphing with SAS/Graph and using PowerPoint for layout (multivariate p-values). When one carries out more than one statistical test, adjustments are needed to keep the overall type I error in check. A number of such methods exist, from simple Bonferroni adjustments to more complicated stepdown procedures and resampling methods. Of interest, then, is the joint distribution between the p-values that arise from statistical tests when the variables lying behind the tests are correlated. Some bivariate p-value densities (pdf approx. 292 k) are shown for the bivariate normal case for several levels of the means and correlation. Of interest (other than the fact that the pictures look really cool!) is that the marginals on Py and Px when the means are zero are still uniform eventhough the joint denisity is highly non-uniform. Since the grid is bounded at 0 and 1, this necessitates the large spikes in the corners. Small multiples and an emphasis on only the data yield another example of "the display is the model": differences in the densities are functions of the rows (correlations) and columns (true means). The text describing the statistics behind what you are seeing is also available (pdf approx. 43k). (All this information was presented at the Joint Statistical Meetings in Minneapolis in early August 2005, along with the results of some small-sample simulations I did with Alok Krishen.) The individual plots were drawn in SAS/Graph and stored as .png files and then pulled into PowerPoint and set on a grid. Text was added as text boxes; the arrows came from 'AutoShapes'. Final rendering was carried out via Adobe Acrobat. Again, PowerPoint shows value for layout, certainly the ability to add text to figures and to line up a grid. But can this tool be easily programmed so as to avoid the mouse-driven manipulation?!? Can one set up text boxes so text can flow one to the next?

Sports ranking scheme (Updated for decommissioning 03 August 2005)

PBStar Golf Course Lots

(Note that in some places above you will need Adobe Acrobat Reader to view pdf files. Of course, you can download this free at Adobe's web site (www.adobe.com).)

Last update: 02 April 2008