2

Imagine a situation where a group of people that are can be characterized as sparse multi-attribute sample (e.g. multicultural MBA class that went to work in many countries) would like to analyze their earnings performance against others. But nobody in the group wants to expose their personal earnings to anyone else, though all are interested in comparison.
How could an experiment be set up in order to:

  • do not expose raw dataset to anyone in the group
  • prevent de-anonymization of each respondent (e.g. we can't allow controlling for country if there's only one graduate working in Country X) Related
  • somehow check that analysis was done fairly (was my input taken into account when averaging)

Everyone posting to sites like glassdoor is not an option, as it will not allow them to distinguish 'their' sample from 'global' data. Only viable alternative seems to involve a mutually trusted 3rd party to confide in and have them perform the analysis.

Are there any other alternatives?

Community
  • 1
Sergii Zaskaleta
  • 71
  • 1
    "Experiment" is a misnomer in this situation. What you're looking for is a data-collection method, I think. – rolando2 Jan 17 '17 at 01:33
  • 1
    My quick reaction: doing this in a decentralized way, without a trusted third party sounds like a difficult operations research, game theory, mechanism design, applied mathematics type problem? – Matthew Gunn Jan 23 '17 at 02:24

0 Answers0