The purpose of editing is to make the data as representative of the real life situation as possible; do this by eliminating omissions and invalid entries, and by changing inconsistent entries. Below are some important principles that should be followed:
- The fewest number of changes should be made to the originally recorded data. You are only trying to make a record or questionnaire acceptable, not make it conform to what you think should be acceptable.
- If you must change a data value, do so only once. If you change a person's age, then later find this age doesn't work for another edit, then you didn't write the original edit correctly. Go back and review the first edit.
- For certain items it may be acceptable to have a "not reported [NR]" or not stated [NS]" category. Thus, in case of an omission or an inconsistent, impossible, or unreasonable entry, a code for "NR" or "NS" can be assigned.
- Obvious inconsistencies among the entries should be eliminated.
- Providing corrected values for erroneous or missing items should be supplied by using other values as a guide; (for example, entries for the housing unit, person, or other persons in the household or comparable group), and always in accordance with specified procedures.
- Specifications for editing the questionnaire data should be developed at the same time as the questionnaire itself. Information to the computer programmer concerning what checks and imputations should be made in the data are provided through editing instructions or specifications. A subject-matter specialist, in collaboration with the computer specialist, should write editing instructions. The instructions describe the action to be taken on each data item. The editing instructions should be clear, concise, and unambiguous, since they serve as the basis for the CSPro editing program. Specifications for machine editing may be written instructions, decision tables, flowcharts, or pseudo-code. Pseudo-code (IF/ELSE logic) is recommended because it can be easily translated to CSPro code. Pseudo-code can be prepared using basic word-processing or text-editing software and can be easily modified.
If the hot deck
method of imputation
is used, it is important that the edit specifications indicate where, during the processing, hot decks are to be updated, that is, at which points in the logic the data items can be considered valid.