Understanding Society User Support

Amelia,

Our weights will be 'optimal' if you have a clear definition of the reference person that's works outside of our study. For example if you define reference person as a head of a household (based on whatever demographical / economic characteristics) this will be just seen as a subgroup in the population, and our weights will represent them as any other subgroup.

I do not recommend using any non-substantive variables for a definition of reference people. For example, adults who answered the household questionnaire. There is no such subgroup in the population, and our weights are not design to represent subgroups that don't exist in the population.

I hope this helps,
Olena

Thank you Olena for your response.

A further question is in studying such representative persons (one person per household, based on demographical/economic characteristics), using a combination of personal level (eg age, gender) and household level information (eg household size, household income), which weights are correct to use, individual weights or household weights? The size of this sample equals the size of the household sample.

Many thanks,
Amelia

Amelia,

Depends on your definition. Is the 'representative person' represents whole households? Or do you want to describe specific people? If the latter you need to have a definition outside of our study, meaning a definition that holds in a society. For example a head of a household (however defined) is a definition in a society, and you would just use individual weights (same as for any other subgroup). This won't be different from studying females, for example. But it is not possible to study people who answered household questionnaire, because such subgroup does not exist in the population. In the latter situation it may be better to study households, or think of a better subgroup definition.

Hope this helps,
Olena

Thank you Olena for your helpful response. The representative person is defined as a head of a household, therefore the size of this sample in my study equals the size of the household sample (1 person per household). I will use data at both the household level (eg household aggregated income) and individual level (eg age) to study the relationship between individual-level variables and household aggregated outcomes, so the correct weight to use is individual weights rather than household weights. Please let me know if my above understanding is correct. Thank you.

Yes, that's correct.
Olena