COMPUTER PROGRAMS FOR ANALYSIS OF LABORATORY AND CLINICAL DATA

点击翻译按钮获取中文摘要

This laboratory continues its tradition of developing novel computer programs for analysis of clinical and laboratory data. A new "universal" curve fitting method has been developed, which combines the advantages of empirical methods (e.g., polyomials) with those of mathematical modelling. The user does not have to provide an equation or mathematical function: The program estimates the curve shape or template automatically by analysis of families of curves. Optimal shifting and scaling factors for both dependent and independent variables are obtained by weight non-linear least-squares curve-fitting with appropriate constraints (e.g., monotonicity, number of inflection points) and sharing of parameters. This method is useful in the context of bioassays, immunoassays, dose-response curves and response-versus-time curves in general. We have also developed multiple new approaches for analysis of episodic hormone secretion in man, experimental animals, and in in vitro perifused cell systems. These methods are statistically valid, objective, reliable, sensitive and yield new physiological information including the instantaneous rate of hormone secretion and the half-life or decay constant(s) for hormone metabolism and degradation. Other programs include improved methods for Lineweaver-Burk and Dixon-plot analysis of enzyme-substrate-inhibitor systems, radioimmunoassays, bioassays, radio-receptor assays, X-ray inactivation and dissociation studies. We have designed and developed a prototype statistical package for microcomputers using macros for popular spreadsheets. This package automatically tests relevant assumptions, assists the user in regard to selection of appropriate tests and options, and assists in regard to interpretation of results.