Here is a sampling of some of the errors, and some ways to eliminate or mitigate them in your study.
Selection error: This is a type of sampling error, in which the interviewer chooses subjects based on accessibility or agreeability, rather than choosing subjects based at random. For example, researchers conducting “person on the street” intercept interviews often select respondents based on those who look most agreeable to being stopped. Similarly, research studies often seek out “friends and associates” first when conducting qualitative interviews, rather than a representative base of respondents.
The solution? Design the study so that a more randomized selection method is used. For intercepts, ensure that interviewers use a formula, such as approaching every fifth person who walks by and meets whatever criteria has been established, irrespective of their appearance or demeanor. For research interviews, create a call grid that maps out the target group and relative percentages of each subgroup, and alternate calls between each of the subgroups.
Questioning error: There are numerous ways in which a research questionnaire can introduce error, such as using ambiguous statements, difficult terminology, or leading phrasing, each of which can lead to respondents’ misunderstanding or unintentional bias to occur.
To reduce or eliminate these types of errors, conduct a test with a small segment of your target group to ensure that the questions are clearly stated, and will elicit the type of information sought.
Changes in tone: Instead of asking a question in a straightforward manner, the interviewer changes his or her inflection. This can unintentionally influence the response, particularly if the issue is a contentious one, where the respondent may be inclined to more strongly agree or disagree, based the inflection of the questioner.
Stress to interviewers the need to ask questions in a similar tone of voice for each interview, and to eliminate leading inflections or offhand qualifiers when asking questions.
Recording error: It’s important to ensure that any recording devices are working properly, and are properly positioned to make a clear recording. If the voices of the interviewer or subject are not clear, it makes transcription far more difficult and prone to errors.
Make sure that all interviewers are familiar with the equipment, and know how to adjust settings to achieve an optimal recording. This usually requires a test run, where the interviewer can hear the results of a bad recording, as well as basic instruction on how to improve the quality (e.g. asking respondents to move closer to the recorder, switching to a more quiet environment, or using the mute button on the telephone when a respondent is answering a question to avoid inadvertent background noise.
Cheating error: Aside from violating research ethics, this type of error is devastating to a research study. Falsifying or changing results will irreparably harm the results of the study, and negates the purpose of doing the research itself.
Proper planning and training of researchers is essential, as well as the use of back checking, can greatly reduce, if not altogether eliminate cheating errors.