Hello, thanks to everybody in advances for your help!
I am working with the MEDW aggregate database, which pools data from 27 surveys of 5 countries, some of national, some of subnational and some of EU elections. I work only with the 11 surveys for national elections.
These include 2 surveys for Switzerland, France, Spain and Germany, and 3 for Canada. These surveys are not for the whole country but each for a given region. Below you can see the regions, as well as the distribution of cases among regions and among countries:
---------------------------------------------------
Cases Percent
---------------------------------------------------
Lucerne 1108 6.4
Zürich 1057 6.1 -> Swiss = 12.5
----
Ile de France 966 5.6
Provence 983 5.7 -> France = 11.3
----
Catalonia 951 5.5
Madrid 976 5.6 -> Spain = 11.1
----
Lower Saxony 975 5.6
Bavaria 4680 27.0 -> Germany = 32.6
----
Ontario 1891 10.9
Québec 1849 10.7
British Columbia 1869 10.8 -> Canada = 32.4
---------------------------------------------------
Total 17305 100.0
---------------------------------------------------
The MEDW includes “within survey weights” useful to restore the sociodemographical (age, gender…) and political (turnout, % votes fore each party) distribution within each region, but does not include “between survey weights” to under- or over-weight data of different regions and countries.
a) Sometimes I want to show and discuss “average”, overall results for the pooled data. How do I do this? Should I weight the cases by region and/or by country? Which weights do you think I should use?
a) Sometimes I want to show and discuss “by country”. But here again it does not seem natural that there are almost 5 observations in Bavaria for each observation in Lower Saxony, to give an example. How do I do this? Should I apply weights? And if so, which weights do you advice me to use?
Again, I’d like to thank you in advanced for your help!
Regards from Madrid,
Andrés
- Badges
- Science topic