As documented in the fieldwork section, reminder letters were sent to all households in which two adults had not responded to the survey.
In any survey involving a postal element (whether invitation and/or completion), there is a delay between a survey being completed and returned by the respondent and its being processed on receipt.
This introduces the possibility of reminder letters being sent to households where the survey has already been completed by one or two adults.
In turn, this can cause duplicate responses, where either a single respondent completes the survey a second time, or where more than two people in a household complete the survey.
As such, there were two potential forms of duplicate response within the final Active Lives survey data: individual-level duplicates and household-level duplicates. Where this occurred, it was necessary to identify and remove duplicate responses from the final dataset.
Because each household was provided with two online logins and up to two paper questionnaires, it was possible for a single household to complete four questionnaires.
Individual duplicates occurred where the same person completed two questionnaires. Household duplicates occurred when more than two different people in the household completed a questionnaire.
As a first step, all cases were identified where the same serial number had been used more than once. These cases were compared, and, where the same age and gender was provided, were assumed to have been received from the same person.
To ensure that as much data as possible was kept, online responses were prioritised over paper responses (since the online questionnaire was more comprehensive and allowed for more sophisticated routing).
After individual duplicates had been removed, responses were checked to ensure that there was a maximum of two responses from adults in a single household.
Where more than two responses had been received, cases were again prioritised by mode and date, and ‘duplicate’ responses removed from the data.
Following this process of deduplication, it was assumed that all responses in the dataset were from unique individuals.
Even after this process, it was possible that unique responses had been provided using the same survey serial number (where two different people both completed the survey using the same serial – one online and one on paper).
For the sake of clarity, these ‘legitimate’ duplicate responses were sorted by mode and date, and renumbered to ensure that each case in the data had a unique ID. These cases have a final digit in the serial number of 3 or 4 rather than 1 or 2.