Survey Error
In survey samples, the primary objective usually is to estimate unknown population quantities, such as population means or totals. The estimates invariably do not exactly equal the population values, i.e. there are survey errors. This article reviews the main sources of survey errors and general approaches to reducing them. Survey errors can be classified in two categories, nonsampling and sampling errors. Common sources of nonsampling error include measurement error, coverage error, mode effect error, nonresponse error, and process error. Measurement error occurs when respondents misinterpret, cannot answer, or inaccurately answer questions. Coverage error occurs when the list used to take the sample does not correspond to all units in the target population. Mode effect error occurs when the way the data are collected influences responses. Nonresponse error occurs when those who do not respond to a survey differ from those who respond to it. Process error occurs when edits, recodes, or other data processing steps go awry. Nonsampling errors are hard to detect and quantify. Their impact can be reduced by careful survey design. Sampling errors occur because only subsets of populations are measured, and estimates vary by sample. Sampling errors arise in every survey, but they are quantifiable.
