search

aaronger / utility-eval-papers

Pipelines and drafts related to arXiv:2312.16201
0 stars 0 forks source link

supplement: consider clarifications r.e. TQ, F_T #66

Closed elray1 closed 8 months ago

elray1 commented 9 months ago

Optionally, there are two things I thought it might be helpful to add to supplement section 4 around the discussion of TQ, $F_T$, etc.:

  1. State that we're using "T" to stand for "total"
  2. Note that $F_T$ is the distribution that arises from summing the marginal quantile sets, but this is not the same as the distribution of the total across all locations. In particular, we are not doing computations here that account for dependence among the locations. Is it the distribution that would arise if we assumed exact dependence among the locations (i.e., something like Kendall's tau = 1)?
aaronger commented 9 months ago

Yeah it is the deterministic dependence you get from making every coordinate a quantile transformation of a single $U \sim \mathrm{Uninf}[0,1]$... not sure how we'd bring a sample stat like Kendall's into this.

aaronger commented 9 months ago

Adding a note about "totalling"...