Cutting-Edge Statistical Methodology

Emphasis on Developing Well-Tested User-Friendly Software

The Division of Cancer Control and Population Sciences (DCCPS) is interested in accelerating the development and application of statistical methodologies for a broader audience of researchers than just statisticians. Supporting the development and widespread distribution of user-friendly, robust software and related documentation for cutting-edge statistical methodology is a high priority for diverse areas of cancer control and population sciences such as surveillance, epidemiology, behavioral, and genetics research. Appropriate statistical methods include, but are not limited to, survival analysis, fixed/mixed/random effects models, Bayesian methods, clinical designs, longitudinal models, and spatial-temporal models.

Software development for statistical methodologies generally moves through three stages of maturity, based on the underlying methodology and the intended user. Stage 1 software, which generally accompanies new methodologies, may be application specific and have inefficient calculation engines. Usage may be limited to the investigators developing the methodology. Stage 2 software should be well tested and the documentation should be easily understandable and usable by other researchers (including some limited use by non-statisticians), although it may have a rudimentary user interface. This or later stages could include collaboration with a clinician or other subject-matter expert to apply the methodology and publish the results in a non-statistical journal. Finally, stage 3 software accompanies mature methodologies where detailed documentation is necessary, and professional programmers provide user-friendly code (with full user interfaces) that is accessible to non-statisticians.

NCI's priority is the movement of code and the accompanying documentation from stage 1, where the developer is the only user, to later stages where it is accessible by non-statisticians. While there are no specific funds allocated for this area, researchers are encouraged to apply to this priority area through various available funding mechanisms announced by the NIH and the NCI. Applicants should demonstrate a need in the broader scientific community for the particular methodology and develop a plan for broad dissemination. The overall goal is to accelerate the time from the initial development of methodology to when it is readily and widely used by statistical practitioners and clinicians.