What determines the appropriate sample size to ensure confidence in research data?
Contrary to popular belief, the right sample size has little to do with the size of the population from which the sample is taken. The most important factor is the selection process used to determine the sample.
Consider this: when measuring the temperature, the sample size is insignificant compared to the surrounding atmosphere it measures. But, you can still measure the temperature of your immediate surroundings. When doctors take blood to determine if you have a disease they don't need to do a blood transfusion, they take just a small sample.
Statistical experts have developed a standard deviation table that can be applied to any size population to determine statistical error. This table expresses error in terms of confidence intervals or variance from the norm.
By using these tables, it's easy to determine the specific sample sizes, which will yield different confidence levels. The tables are based on "like" populations sampled at random. This table, when translated to plain English reads as follows:
These sample sizes and errors apply to any population larger than the sample. If you draw a sample of 400 from a population of 1,000, 2,000, 4,000, 40,000, 400,000 or 240 million, the error will remain within plus or minus 2.5%. Example: If 65% of the respondents say "yes," the true answer would be between 67.5% and 62.5%.
The results from your sample represent the thinking of the entire population only if the sample is drawn in a proper way. In most cases you'll need to work from a list of names (customers, residents, registered voters, members, etc.) A computer is usually used to generate random sample lists.
First -- divide the number of interviews you wish to conduct, for example 400 for a 5% error margin, into the approximate number of people on your list. This is an interval, and it's usually 10 or less, unless your list is very large. Let's say the interval is 10 for now.
Second -- select a starting number by chance from one to whatever your interval is. Let's say you pick seven. The seventh person on your list would be your first respondent. You would then interview every tenth person from that starting number - 17th, 27th, etc. You now have a representative sample of the population. The thinking of those interviewees will more confidently represent the thinking of the larger population.
For a more targeted approach, samples can be selected from two or more distinct population groups. For instance, comparing existing customer responses with targeted prospect responses. This approach will help a company not only find out what their customers think, but will help determine what will best persuade prospects to become customers.