Medical research frequently involves longitudinal and/or multivariate measurements; when the research involves humans or laboratory animals, the data are often incomplete or involve time-dependent covariates, problems that traditional statistical techniques and software cannot handle properly. Many important advances have been made in statistical methods for incomplete longitudinal/multivariate data in the past 5 to 10 years, but these methods are not generally available to medical researchers due to a lack of highquality, user-friendly computer software.The ultimate goals of this project, including Phase II, are to produce and market such software for use on mainframes and large microcomputers. The software will implement methods for linear models with linear covariance structure, subject to linear constraints, which have been shown to be extremely useful for analysis of incomplete medical research data. Phase I will include research to: (1) select algorithms for the fundamental computations; (2) design overall program structure; (3) design two user interfaces (menu and command-language driven) for each program; and (4) implement, debug, test, and document prototype software. Phase II will include: (1) research with the prototype software to improve user interfaces and program performance and (2) development of highly efficient, user-friendly production software which will be marketed.Division of Research Resources (NCRR)
