Annual sample weights
Weighting is required to reduce the bias in survey estimates and weights are produced to make the weighted achieved sample match the population as closely as possible.
For the Active Lives survey, the weights correct for the unequal selection of addresses across local authorities and for the selection of adults and youths within households. They also adjust the achieved sample by month to control for seasonality.
In addition, the sample is weighted to match ONS mid-year population estimates and national estimates derived from the Annual Population Survey.
The data need more than one set of weights to take account of the following elements of the design: postal and online data and the need to analyse data in a time series.
The final data include eight sets of weights:
- wt_final: for data collected in both the online and postal questionnaires;
- wt_final_online: for data collected in the online questionnaire only;
- wt_final_B: for data collected among participants assigned to the online group 1;
- wt_final_C: for data collected among participants assigned to the online group 2;
- wt_final_AB: for data collected in the postal questionnaire and among participants assigned to the online group 1;
- wt_final_AC: for data collected in the postal questionnaire and among participants assigned to the online group 2;
- wt_time: for analysis of data collected in both the online and postal questionnaires when analysing by month or quarter; and
- wt_online_time: for analysis of data collected in the online questionnaire only when analysing by month or quarter.
The same weighting strategy was used for each set of weights (this strategy is described below).
Note that an approach that weighted each module separately and then combined them would have been less complicated.
However, this approach was not an option as the postal sample is issued after the online. The postal sample is therefore not nationally representative and so cannot be treated as a separate nationally representative module.
Quarterly and monthly weights
We defined months and quarters by when the questionnaire was completed/returned (rather than issued) and we used survey year months and quarters rather than calendar quarters (see the data validation chapter for more detail on this).