Sample Size 

Too small a sample will give unreliable results, but too large a sample wastes money and time. Large samples can also lead to less reliable results since it is difficult to supervise closely the collection of a lot of information. The correct sample size is determined in consultation with a statistician. Practical considerations such as funding, time and manpower, the sample size as determined by statistical formulae and the proposed study are all taken into account when deciding on the sample size. 

The following must be considered before a sample size can be determined: 

• The expected size of the population which will be investigated (thousands or only tens).  
• The expected prevalence of a characteristic that will be studied or the size of the difference between two groups (for example means or percentages), based on experience, a previous pilot study or knowledge of the literature. 
• How reliable the results should be with regard to extrapolation to the population. If the expected prevalence is 30% and the researcher wants to be 95% sure that the sample results are within 5% of the true population value, a sample of 323 is needed. If the researcher wants to be within 10% of the population value, only 81 subjects are needed. 
• The number of variables or subgroups (strata) which must be investigated. A subgroup is a subsection of a population with unique characteristics. If the population is the rugby playing primary school boys in Bloemfontein, different age groups, races, schools or positions which they play in would constitute subgroups. If the study must provide results per subgroup the sample size per subgroup must be large enough to obtain reliable results. 
• The availability of resources (manpower, time, funds). Here one should also estimate how long it will take to obtain the relevant information (for example through interviews, laboratory investigations, or the follow up of patients).