DESIGN, TESTING AND ESTIMATION IN COMPLEX EXPERIMENTATION. III. THE DESIGN AND ANALYSIS OF MULTIVARIATE SENSITIVITY EXPERIMENTS.
Final rept. for period ending Feb 65,
ROCKETDYNE CANOGA PARK CALIF
Pagination or Media Count:
This report contains research results obtained on the design and analysis of quantal-response experiments involving one or more stimuli. In particular a univariate nonparametric design for estimating the stimulus level corresponding to a given response-fraction, alpha, is developed. Special attention is given to extreme values of alpha. In addition, an algorithm for determining maximum-likelihood estimates of partially ordered probabilities and its application to various aspects of multivariate sensitivity experiments are described. A method for handling such experiments when the stimuli do not interact is discussed and a nonparametric statistical design with point estimates is given. Finally, response models are developed for various conditional distribution requirements on the stimuli. For a large class of these response surfaces the equal-probability response contours are shown to be hyperbolas. Author