INTERVAL ESTIMATES FOR RATIOS OF REGRESSION PARAMETERS.
Abstract:
Two extensions of the classical estimation of a ratio of two regression parameters are considered. The first extension is concerned with the case in which a number of experiments are performed and each experiment involves a regression function with different regression parameters and different unknown variances of observation. With such assumptions it may turn out that even though the regression parameters for each experiment are different, the ratios of two specified regression parameters for each experiment are the same for all experiments. The first extension provides a method of combining the results of all experiments to obtain an interval estimate for this unknown ratio. The second extension provides a solution to the problem of obtaining simultaneous interval estimates for more than one ration of regression parameters when a single experiment is performed. Author