When subjects for a study are selected, researchers only need to take into account the characteristics of interest to the study. In the example of the ER, factors such as age, income level, education, insurance, and health status are relevant. Suppose that in the area the hospital serves, 5% of the population has high incomes and 10% of the population is over 65 years of age. If the sample has similar proportions of subjects with these and other relevant characteristics, it is a representative sample. Height, weight, color of eyes, state of origin, and hundreds of other such characteristics are not relevant to the study and need not be considered.
Careful selection of variables is important to keep the sample size manageable. There may be 50 variables to describe patients’ characteristics such as age, ethnicity, gender, average salary, annual household income, self-reported health status, dietary habits, number of children, and on and on. The norm is to have 20–30 subjects per variable. If half of the 50 variables are included, the sample size may go up to (25 x 20 =) 500.
Suppose the research is to be done through survey questionnaires. A typical projected response rate for a mail-in questionnaire is 30% if there is an existing relationship with respondents and 10% for “cold turkey” marketing surveys. So, for a sample size of 500—500 responses—you will have to send questionnaires to a number of people, X, such that 500 is 30% of X. To get 500 completed questionnaires, the researcher would need to send out at least 1,667 questionnaires! To design a practical study with a practical sample size, researchers have to clearly define their variables and decide which of them are essential and which are not.