| Sumario: | This article examines the validity of schedule entries in interviews and surveys as of September 1936. Data secured by interview or survey are subject to three major classes of error: inaccurate or biased responses of the people interviewed, misinterpretation or inaccurate recording by the interviewer or enumerator, and sampling error. The first of these classes can be minimized by the use of simple and objective questions, or by internal schedule checks. The third is usually estimated from probability theorems. The second is frequently ignored. Errors of interpretation and schedule entry are especially important, however, and doubtless have been overlooked in many studies that were thereby rendered actually invalid. Records of quantitative reports, or 'yes' and 'no' statements can be expected to have high accuracy. It is concluded, therefore, that some instruction of recorders increased the proportion of agreement by approximately 10 percent. It is apparent, therefore, that workers sent out for survey work or interviewing must be carefully drilled. The fact that a small amount of instruction can influence so markedly the validity of schedule entries merits careful attention during a period when many hastily organized and rather extensive surveys are being conducted.
|
|---|