Data Processing

Data Entry

The data from the Household Form, Household Questionnaire and Person Questionnaires were manually entered into a database. The keyed numerical data were subject to 100 per cent verification (i.e., the data was entered in twice and any discrepancies corrected). The keyed verbatim responses were only entered once as these were only used for coding purposes and any mis-entered data could be easily identified and corrected. During data entry, the data was checked using range, logical and consistency edits. Where necessary the data entry was suspended until the identified problem was resolved.

The data from the Self–Completion Questionnaire were scanned into a database using a mark-sense recognition system. A sample of forms, together with forms that had a high proportion of missing responses, were visually inspected to ensure the scanning process was working correctly. All inappropriately marked multiple responses were also visually inspected, and where no single response could be clearly differentiated from the scanned image, the following rules (based on the SF-36 rules for coding problems in Ware et al, 2000) were applied:

• If a respondent marked two responses that were adjacent to each other that were part of a scale, one was randomly chosen.
• If a respondent marked two responses for an item and they were not adjacent to each other, the item was coded as -5 “invalid multiple response”.
• If a respondent marked three or more responses for an item, the item was coded as -5.

Once the data from all forms were entered, consistency in the data recorded on the various forms was checked. Any discrepancies were then investigated and resolved.

Coding Responses

The coding of the occupation and industry questions was done in the office prior to data entry. Occupation questions are coded to four-digit Australian Standard Classification of Occupation (ASCO) and industry questions are coded to four-digit Australian and New Zealand Standard Industry Classification 1993 (ANZSIC).⁴⁹

The questionnaires also contained items for which a partial list was provided with an ‘other, please specify’ category and some of these lists needed to be further extended. These codeframe extensions occurred once a sufficient number of forms had been entered so that common responses could be identified. Some backcoding into the original codeframe in the questionnaire was also done where the interviewer had written down an ‘other’ response that actually fell into the codeframe already provided in the questionnaire.

The accuracy of the coding was monitored on a continual basis. Senior coders verified 10 per cent of the occupation and industry codes. Any discrepancies identified were discussed between the coders and corrected, thus identifying problematic areas for further investigation and providing feedback to the coders. The coding of the ‘other, please specify’ responses were checked through a blind re-coding of all cases. Wherever problems were identified, these were investigated and resolved.

Endnotes: