Developing Methods for Complex Sample Surveys

EIFACS (Expanded Integrated Facility for Analysis of Complex Surveys) will provideepidemiologists at NIAAA and other researchers with the capability to perform valid multivariatestatistical analyses on survey data involving complex sampling designs in an easy-to-use windowed PCenvironment. Variance estimation will utilize Taylor series linearization, jackknife, and, possibly,bootstrap techniques. Essentially any practical sampling design can be accommodated (up to seven nestvariables). Analysis procedures envisioned include Weighted Kappa, Fourfold Tables (including tablestratification), discriminant analysis, survival analysis, cluster analysis, and structural equation modeling.The effort capitalizes on the existence of IFACS (Integrated Facility for Analysis of Complex Surveys),which has taken major steps in the ultimate direction planned for EIFACS. Phase I results will includealgorithmic development of the various new or augmented procedures, along with their varianceestimation methods; a Monte Carlo test facility (to verify validity/reliability of the procedures andvariance estimation); and improved design of the windowed interface for user navigation through theavailable input, procedure execution, and output functions. Phase I output will provide a clear definitionof all new and modified procedures at the algorithmic level for actual implementation into computer codeduring Phase II. Phase I will also provide a partial demonstration of a procedural augmentation togenerate confidence in the ultimate success of the effort.