Please activate JavaScript!
Please install Adobe Flash Player, click here for download


ISCB 2014 Vienna, Austria • Abstracts - Oral Presentations 21Monday, 25th August 2014 • 14:00-15:30 Monday25thAugustTuesday26thAugustThursday28thAugustAuthorIndexPostersWednesday27thAugustSunday24thAugust Monday, 25th August 2014 – 14:00-15:30 Invited session I2 Beyond R packages: getting our methods into standard software Organizer: Georg Heinze I2.1 Writing and developing statistical software: the statistical methodologist´s view IR White1 1 MRC Biostatistics Unit, Cambridge Institute of Public Health, Cambridge, United Kingdom This talk describes my experience of writing and developing statistical software. Much of it has been universally ignored, but some of it has been widely used. The key to wide uptake is to tackle an important general problem rather than to focus on a specific solution. I will discuss the importance of choosing a syntax which is both easy to use and flexible, and of using the software extensively oneself (“what happens if I try this?”). Most important in my experience is establishing two-way communications with users, since no developer can anticipate the range of uses to which a good piece of software will be put. Such two-way com- munication also often raises interesting methodological issues. I will illustrate these ideas using my experience of developing software in Stata for multiple imputation (with Patrick Royston) and for multivariate meta-analysis. The former was ultimately superceded by a package in “of- ficial”Stata, which reproduces much of our functionality. I2.2 Beyond wild horses: developing innovation at Cytel Y Jemiai1 1 Cytel Statistical Software and Services, Cambridge, United States Henry Ford once said, “If I’d asked customers what they wanted, they would have told me,‘A faster horse!’” With the power and flexibility of R, statisticians all over the world have been able to breed their very own“horses,”but how does one make a“car?” This is what we do at Cytel.To develop innovative statistical software prod- ucts that people love, we ask ourselves three key questions: 1. Of all the new statistical methods being proposed on a daily basis, which ones will be useful to the statistical community? 2. How do we create an exceptional user experience that helps spread the use of innovative statistical methods? 3. How do we establish trust in the results produced by the software? Answering these questions is a complex undertaking. The process by which we answer these questions relies on a combination of frequent in- teractions with our customer base, strategic consulting engagements to solve our clients’ most pressing and challenging statistical problems, and internal methodological research. This talk describes how Cytel develops unique and innovative software products in an effort to serve the statistical community and promote the application of better statistical techniques in the scientific community. I2.3 An inside perspective on the development of SAS statistical software RN Rodriguez1 1 SAS Institute Inc., Cary, United States This presentation provides a behind-the-scenes look at how SAS/STAT software is developed, beginning with the qualifications we consider in recruiting research statistician developers, and progressing through the stages of design, programming, testing, documentation, and user sup- port. Our decisions about which methods to implement are based on cus- tomer requirements, which are driven by the increasing value, complexity, and size of data, coupled with advances in methodology and technology. These decisions are also informed by discussions of promising methodol- ogy between developers and researchers. In order to deliver versatile methods that work robustly across many areas of statistical practice, developers must often extend the literature on the- ory and methods to handle issues such as unbalanced data, large numbers of random effects, and missingness. Again, this presents opportunities for interacting with researchers. Our development process also emphasizes syntax that is common across procedures, clear and consistent output, high-performance computing, numerical accuracy, and helpful documen- tation. Examples drawn from recent releases illustrate these principles.

Pages Overview