It is often useful to weight within time periods. For example, if a survey is repeated every quarter, it is usually appropriate to weight each quarter separately. This can be done either by constructing the weight within a sub-sample of the data or by using the time periods as a variable. This is done as follows:
- If it does not already exist, create a categorical variable that represents the time periods.
- Use this categorical variable as an adjustment variable.
- Set the targets to 1/k, where k is the number of time periods. So, if you have four waves, then each target is 0.25 (it is also possible to do other things, but your analysis will become complicated).
- Create a new set of all the other adjustment variables for each time period, recoding the observations not in that time period to 0. For example, if you are weighting by a composite age and gender variable with eight categories, and you have four time periods, then you need to create 32 variables.
- Divide each of the target values by For example, if the target for Females aged 18 to 24 is 0.1 in the total population, then the target set for each of the four time periods is 0.025.
Please sign in to leave a comment.