OPTIMAL ALLOCATION OF OBSERVATIONS WHEN COMPARING SEVERAL TREATMENTS WITH A CONTROL.
CORNELL UNIV ITHACA N Y DEPT OF OPERATIONS RESEARCH
Pagination or Media Count:
The problem of making one-sided comparisons of p treatment means with a control for normal populations is considered. An exact solution is given for optimally allocating observations between the treatments and the control for the case in which the p treatment populations have a common known variance which may be different from the known variance of the control population. The solution is given in the form of an equation for or 2 the equation is solved numerically for p 2, and a table of these solutions is provided. This problem was considered by Dunnett who proposed an approximate solution which was to be used when the confidence coefficient is approximately 0.95 or greater. It is shown, as a corollary of our main result, that his solution is correct asymptotically as the total sample size available for experimentation is allowed to grow arbitrarily large. Author
- Operations Research