Microsoft Word research method fm doc

 
Sample size calculation using the Epi Info (STATCALC) program: For population 
survey or descriptive study 
 
Assumptions: 
1) 
The sample to be taken must be a simple random or otherwise representative 
sample. A systematic sample, such as every fifth person on a list, is acceptable 
if the sample is representative. 

---

Research methodology 
114
2) 
The question being asked must have a "yes/no" or other two-choice answer, 
leading to a proportion of the population as the final result. 
Example
: Suppose that you wish to investigate whether or not the true prevalence of HIV 
antibody in a population is 10%. You plan to take a random sample of the population to 
estimate the prevalence. You would like 95% confidence interval that the true proportion in 
the entire population will fall within the confidence intervals calculated from your sample. 
In STATCALC, therefore, you enter the population size, say 5000, the estimate of the true 
prevalence (10%), and either 6% or 14% as the "worst acceptable" value, the end point of 
your sample confidence interval. The program then shows the sample size for several 
different confidence levels, including the 95% we desired. 
The "worst acceptable" value is one of the confidence limits around the estimated sample 
proportion. The sample size given is for a "two-tailed test", a larger sample size than for a 
"one-tailed test". The equation works with the following values, if the confidence interval is 
95%. 
Upper worst acceptable = proportion (P) + (1.96 standard error) = P + marginal error 
Lower worst acceptable = proportion (P) - (1.96 standard error) = P - marginal error 
 
In the sample size calculations, an initial screen explains the data items and allows input of a 
set of values. Pressing then shows the results (calculated sample sizes) on the second 
screen. 
Sample size calculations for different study designs (more complex designs) are also 
provided by the STATCALC program.