Closed elray1 closed 8 months ago
Actually I feel like this is more like "as known from folklore"... and is useful to write out because there is no good reference that succinctly pulls these things together. Neither Dawid nor Gneiting seem to have ever wanted to bother. I'm heavily influenced by Rockafellar and co these days, but their language is pretty quirky...
it seems like we should somehow concretely indicate that this is a longstanding known fact... at minimum, I think Murphy(1993) has some kind of statement of this
But that's for a discrete problem... It's the continuous situation that's hard to reference. I wish Ehm et al. (2016) had done this in sec 2.3, but they really didn't.
That makes sense. Can we say something about how for discrete problems this is well known (e.g., cite), and it seems to be understood as true in a continuous setting as well though we are not aware of a specific citation, so we lay out out here?
I'm torn. On the one hand, I don't want to look like I'm trying to recycle content that Dawid or Gneiting would say is common knowledge as original, but I also feel like the presentation and level of detail might be novel. There is the stochastic optimization literature, but will take some time to find a sufficiently non-quirky source...
I'm not saying you have to attribute your work to someone else, but I think we need to do something to say how the content of this section fits into or complements what's out there in existing literature
How about Gneiting "Quantiles as optimal point forecasts" and/or Jose Winkler "Evaluating Quantile Assessments" along with a "see references therein"...
sounds good to me
first try: bf5bb2a13f9982e50f4d6a94c3a6289d06800721
In supplement section 3, "[w]e recall how quantiles arise as solutions to a probabilistic decision problem." Let's add a citation to the literature that we're recalling.