WELL? WHAT DO YOU THINK?
ANNEX B. SURVEY ADMINISTRATION
In advance of fieldwork, a letter on Scottish Executive headed paper was sent to the "Occupier" at all sampled addresses - the version of this document used for the main sample is reproduced in Annex C to this report. All fieldwork on this project was conducted by fully-trained interviewers from the NOP fieldforce, working to the criteria of the Interviewer Quality Control Scheme. Each address received at least six calls (including at least two in the evening or at the weekend) before it was treated as a non-contact. Contact Sheets were returned to NOP while interview data was downloaded via modem from the CAPI machines. Information on number of households and individuals was recorded onto the CAPI system for use at the weighting stage. A total of 1381 interviews were conducted with the main sample between late July and early September along with 51 interviews on the ethnic minority booster sample. The contact rate details are shown in Table B.1.
Table B1
Main survey | N |
Issued sample | 2880 |
Less invalid addresses (non-residential) | 159 |
Remaining sample | 2721 |
Successful interviews | 1381 |
Refusal before respondent selection | 430 |
Refusal after respondent selection | 235 |
Not available after 6+ calls | 485 |
Too ill | 36 |
Away during fieldwork | 43 |
Needed mother tongue interviewer (not possible inside fieldwork period) | 9 |
Withdrawn (interviewers threatened) | 52 |
Other outcomes | 50 |
Minority ethnic booster survey | N |
Issued sample | 480 |
Less invalid addresses (non-residential and white-only households) | 269 |
Remaining sample | 211 |
Successful interviews | 51 |
Refusal before respondent selection | 12 |
Refusal after respondent selection | 8 |
Not available after 6+ calls (likely to include some more non-eligible households) | 85 |
Too ill | 1 |
Away during fieldwork | 4 |
Needed mother tongue interviewer (not possible inside fieldwork period) | 12 |
Other outcomes | 38 |
The main survey data was weighted firstly by differential chance of selection for interview and then by the most recently available demographic profile supplied by the Scottish Executive. The effective sample size was calculated as 79.8% of the total figure, i.e 1102.
The resulting maximum sampling errors (on a finding of 50%) are shown in Table B2, both for the total sample and for some sub-samples. The errors also allow for the estimated impact of clustering the sample as described in the previous section (a factor of 1.25). The target sample size for the main survey was 1500 and the smaller achieved sample meant that the quoted errors are slightly larger than they otherwise would have been (e.g. the maximum error on the overall sample would probably have been +/- 2.5%, rather than +/- 2.7%).
Table B2
1381 = 1100 effective (i.e. the total main sample) | +/- 2.7% |
438 = 350 effective (e.g. least affluent areas) | +/- 6.5% |
225 = 180 effective (e.g. saw the male depression vignette) | +/- 9.1% |