Key points are not available for this paper at this time.
In planning a sampling investigation it is desired to know a certain standard deviation to within 10 per cent with a confidence coefficient of 95 per cent. How many observations do we need in order to achieve the required accuracy of estimation? This paper provides an answer to this and siinilar problems. Given n observations from a normal distribtution with mean , and variance a-2, then fs2/o-2 has the Chi-square distribution with f=n -1, degrees of freedom. s2 is the usual unbiased estimate of (J2. In its turn, for large f, 2-x follows the normal distribution with mean 2f and variance 1. A more accurate approximation is obtained by taking the expectation of 2x2 to be /2f-, as in Fisher and Yates Tables, for example. However, the use of -\'2f seems to lead to a satisfactory degree of closeness for small f and is simpler to use. Hence,
Thompson et al. (Thu,) studied this question.