Data collected for compiling statistics frequently contain obvious systematic errors; in other words, errors that are made by multiple respondents in the same, identifiable way (see “Statistical Data Editing – Main Module”). Such a systematic error can often be detected automatically in a simple manner, in particular in comparison to the complex algorithms that are needed for the automatic localisation of random errors (see the method module “Statistical Data Editing – Automatic Editing”). Furthermore, after a systematic error has been detected, it should be immediately clear which adjustment is necessary to resolve it. For we know, or think we know with sufficient reliability, how the error came about.
A separate deductive method is needed for each type of systematic error. The exact form of the deductive method varies per type of error; there is no standard formula. The difficulty with using this method lies mainly in determining which systematic errors will be present in the data, before these data are actually collected. This can be studied based on similar data from the past. Sometimes, such an investigation can bring systematic errors to light that have arisen due to a shortcoming in the questionnaire design or a bug in the processing procedure. In that case, the questionnaire and/or the procedure should be adapted. To limit the occurrence of discontinuities in a published time series, it can be desirable to ‘save up’ changes in the questionnaire until a planned redesign of the statistic, and to treat the systematic error with a deductive editing method until that time.
To read the entire document, please access the pdf file (link under "Related Documents" on the right-hand-side of this page).
Your feedback is appreciated. Please send your remarks, suggestions for improvement, etc. to email@example.com.