Example 1: Someone claims that US females are more
likely than US males to study French. Our null hypothesis
is that the proportion of female French students is 0.5. We wish
to compute the sample size that will give us a 80% power to
reject the null hypothesis if the true proportion of female
French students is 75%.
The values 50 and 75 should be entered in the fields
"Proportion 1" and "Proportion 2", respectively, 80 in the field
"Power", the ratio and alpha risk may be left to their default
values. The check box "One-sided test" should be left unchecked, but
the check box "One-sample test" should be checked because the
null hypothesis is that the percentage of females is 50% in
the population. Sampsize returns a sample size of n = 29.
Cluster sampling:
The sampled population will be drawn from 8 universities. We know that
students within a given university are more alike than students from
different universities. We suspect that the intraclass correlation is
fairly high (0.1), and wish to correct the above sample size to take
into account the sampling design. We check the "Cluster sampling design"
check box, enter 0.1 in the "intraclass correlation" field and 8 in the
"number of clusters" field. Sampsize returns a corrected sample size of
n = 96, with an average of 24 students sampled in each university.
Example 2: We want to conduct a survey on people's opinions of
the President's performance. Specifically, we want to determine
whether members of the President's party have a different
opinion from people with another party affiliation. We estimate
that only 25% of members of the President's party will say that
the President is doing a poor job, whereas 40% of other parties
will rate the President's performance as poor. We compute the
sample size for alpha = 5% (two-sided) and power = 90% by
entering the values 25 and 40 as "Proportion 1" and "Proportion
2", respectively, 90 for "Power", "Ratio" and "Alpha" risk are
left at their default values (1 and 5). We leave both check
boxes unchecked. Sampsize returns an estimated sample size of
216 for each groups.
|