Vormistamine ülesanne 3
inaccurate, departs from the “true” value (see also Hox, Chapter 20).
A perfect survey would minimize all four sources of errors. Coverage error is avoided when every
member of the population has a known and nonzero chance of being selected into the survey.
Sampling error is reduced simply by sampling enough randomly selected units to achieve the
precision that is needed. Nonresponse error is avoided if everyone responds or if the respondents
are just like the nonrespondents in terms of the things we are trying to measure. Measurement error
can be prevented by asking clear questions; questions that respondents are capable and willing to
answer correctly. In the survey design stage the methodological goal is to prevent or at least reduce
potential errors; in the analysis stage the statistical goal is to adjust the analysis for errors in such a
way that correct (i.e., unbiased and precise) results are produced. The methodological survey
annaabi.ee 
