I have a relationship database of circa 11,000,000 observations. These 11,000,000 observations relate to circa 4,000,000 individuals. Initially the idea was to group these individuals into households depending on their relationships. The issue I'm noting now is that these relationships are not mutually exclusive.

For instance, take 4 individuals:

  • Person 1 is linked to Person 2
  • Person 2 is linked to Person 3 and Person 4
  • Person 1 is linked to Person 3 but not Person 4.

Trying to assign unique household ids to these individuals will result in people being put into different households etc.

What I was thinking was trying to use some kind of social network analysis and weighting system to assign people into houses.

The packages I have at my disposal are SAS and R.

I'm wondering has anyone any links to papers which have attempted something like this before?

  • 1
    $\begingroup$ Just to avoid recreating the wheel, did you look into their ID number? Some studies incorporate household ID into the generation of ID number, and you may not have to do all these works. Also, overall, the challenge here is not clear. What is "linked" and what is a household? For instance, if Person 4 is Person 1's mother in-law living else where, do you intend to call them one household? $\endgroup$ – Penguin_Knight Oct 9 '18 at 14:34
  • $\begingroup$ The idea was to design a criteria to weight each relationship. If they're married, if they're within the same age etc. I would then form a weighted adjancey matrix and have threshold of 0.8 or some number as to having a strong link. $\endgroup$ – Sean_C Oct 9 '18 at 15:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.