search

sdcTools / recordSwapping

c++ code for rank-swapping
0 stars 1 forks source link

question regarding the data (individual or household data?) #2

Closed pnicky closed 5 years ago

pnicky commented 5 years ago

Hello, I'd like to try this method. The test data are running fine. What I don't understand is, do we use individual data (eg. sex, age), or household data (eg. household size). It seems like it is combined somehow. Do we have to have 4 geography level? Does the data have to be grid-like? Thank you for that! Niki Pálmai (HCSO)

alexkowa commented 5 years ago

Hi Niki, each row of the data set should be an individual, however the individuals are connected into households with a household id. the number of hierarchy levels in the geography is flexible, it can be simply controlled with the number of variables provided as geography variable.

Alex

pnicky commented 5 years ago

Thank you, Alex! And do you specify the similarity of two households based on both household and individual attributes? Does the method take into consideration each member of the household? (I think in Japan only the head of household matters according to their methodology.) Do you know if anybody else tests the code too? Thank you again! Niki

alexkowa commented 5 years ago

The similarity is computed based on household attributes only, but of course in this way information on the head of the household can also be used. We collected the information on the testing of the prototype here: https://github.com/sdcTools/protoTestCensus not sure if you have seen that. We have tested it of course within the project and the prototypes were presented at the WG Census with the hope to find additional testers.

pnicky commented 5 years ago

Thanks again. Yes, I've got this doc. I'm going to test this further soon hopefully.