Friday, April 18, 2014

Good post on Vox about issues with ordinal variables

This was a very good article tackling the issues of trying to assign an ordinal score to a multi-dimensional variable.  Mark has been saying this for years, already, but it is good to see statements like this coming out of more mainstream groups:
The problem with ordinal rankings — and the more variables, the more problems here — is that it implies meaningful differences between one job and the next one that is one ranking below it. You can definitively say that one job pays more than another, but is it true that clinical social worker is better than nail technician is better than middle school teacher, as US News' rankings imply? And even if somehow that were empirically provable, what's the practical application of this knowledge? Should the middle school teacher go be a social worker?
Rankings lists can occasionally provide useful functions but it is good to see more discussion of the limitations of these measures.  Now who is brave enough to do this with post-secondary education?

No comments:

Post a Comment