2.3 Sample Size
Experience teaches us that the responses of a large sample tend to balance out and better represent the population of all potential respondents, whereas the responses of a small sample (i.e., less than 30 respondents) tend to skew. Thus we can only analyze a subgroup of a sample if we expect at least 30 reponses. For example, when 150 unbiased respondents answer a question with 5 possible answers, we expect 30 responses for each response option.
We can predict the population's responses when we analyze at least 30 responses from a proportionally representative sample.
Table 5. Validity Errors
|Types of error||Definitions||Examples|