You are on page 1of 1

When one uses the results of a random sample to estimate the value of a population

proportion P, often one is asked to determine a sample size n such that the one is x% confident
that the sample proportion p is within c of the (unknown) population proportion P. The
required sample size depends on four factors, the precision required (c), the confidence level
required (x), the population proportion (P), and the size N of the population from which one is
sampling. The required sample size is given by the formula

where zx is the upper x% point of the standard normal distribution. Following are some common
values of x and zx
x
zx

80% 90% 95% 99%


1.2816 1.6450 1.9600 2.5758

When one has some sense of the magnitude of P, this formula produces the required
sample size. When one has no idea of the magnitude of P, one can use the value P=0.5 to
produce a conservative sample size, large enough to guarantee the precision at the given level of
confidence for any value of P. The conservative value of n is given by

For large values of N the second term in the brackets is close to 0, and so this expression
for n reduces to

Note also that the value of zx when x is 95% is 1.96, or approximately 2.0, so that in this case
this formula reduces further to

Thus if the required precision is c=.05 one needs a conservative sample size of 400. This simple
formula also tells us what the precision is, at 95% confidence, for any sample size. It is

You might also like