1 Introduction | 2 Sampling methodology | 3 Questionnaire development and testing | 4 Fieldwork methods and interviewers | 5 Checks and audits | 6 Response rate and interview length | 7 Classifications and coding | 8 Survey weights | 9 Imputation | 10 Variance estimation and significance tests | References | A1 Response rate by interviewer experience | A2 Sample and population profiles | A3 ACNeilsen area sampling frame | A4 Effect of area unit population changes | A5 Derivation of eligibility probability estimate | A6 Investigation of incident dates | A7 Contact sheets | A8 Showcards | A9 Selected CAPI screenshots
CAPI range and consistency checks
Several range and consistency checks were embedded in the CAPI questionnaire script to ensure the data was of high quality. For instance, if the value of a stolen vehicle was reported as being $50,000 or more at Q85 a pop-up window appeared, prompting the interviewer to check that this amount was correct.
Interview checks and validation
Checking and auditing of interviewers’ work was undertaken at two levels. This was through (i) area coordinators’ checks, and (ii) telephone validation interviews.
Checks by area coordinators covered fieldwork support material - i.e. all area maps, call sheets, respondent selection sheets and analysis forms. Area maps were checked to ensure that all designated dwellings were within map boundaries, that the correct start point was used, that the correct calling procedure was used after the start point, and that the call address record form was completed. Call sheets were checked to ensure that the calling and listing procedures were correctly followed. They were also checked against timesheets. Respondent selection sheets were checked to ensure that the correct respondent was selected.
Telephone audit interviews were conducted for a systematic selection of 25% of each interviewer’s work, spread over each NAU the interviewer worked in. These audits confirm that the interview took place according to the prescribed standards. Originally ACNielsen planned that where an interview selected for audit was obtained from a dwelling with no telephone, a mail audit would be undertaken. But on later consideration it was decided that a mail audit was inappropriate for this survey given the confidentiality and safety issues this raised, so no audit interviews were conducted by mail. Sufficient dwellings with telephone numbers were available to adequately check each interviewer’s work. If this had not been the case for any individual interviewers, their work would have been checked through a door-to-door audit.
Due to the sensitivity of many questions in this survey, the audit information centred on length of interview, time of interview, and respondent selection procedures (rather than repeating questions from the main questionnaire).
These audits revealed that one interviewer had been falsifying interviews. The affected areas were reworked, which delayed the end of the fieldwork by one week.
ACNielsen checked the plausibility of frequency tables and their compliance with skip instructions. They also ran a set of programmed edit checks in SPSS. This uncovered only a small number of irregularities.[12]
More detailed checks on consistency naturally arose during more detailed statistical analysis by Statistical Insights. These checks identified a problem with coded Victim Form data, which arose from a known bug in Confirmit. They also identified another 49 cases with inconsistent data. Most of these occurred early in the Victim Form.
Some possible outliers were apparent at the incident screener questions. These were examined to see whether they showed any signs of respondents being unduly creative, but most did not seem implausible. Their treatment is described in Chapter 9 on imputation.
Footnote
12 The most common (affecting 37 cases) was corrected by reference to contact sheets. It arose from the CAPI validity check programmed for Q147 concerning the number in household aged under 15 years being imprecise.