On the Determination of Sample Sizes in Designing Experiments

Methods are developed for determining sample sizes required to estimate means of normal populations with given precision, or to give significance when means differ by a specified amount. The methods assume that an estimate of the variance has been obtained from a previous or preliminary experiment. A simple device is suggested for using a priori information about the variance when a formal estimate is not available. * Most of this work was done under contract N7onr371 with the Office of Naval Research.