« first day (5011 days earlier)      last day (292 days later) » 

00:24
@Sycorax That's an interesting problem; classifying too early vs waiting too long. Have you found any applicability of optimal stopping theory to this tradeoff?
 
1 hour later…
01:27
@Galen the term of art is early classification. There’s a bunch of different ways to do it.
 
5 hours later…
06:56
@Sycorax: reminds me of the story of the car mechanic that gave the malfunctioning engine one solid THUMP with his hammer, whereupon the engine ran again and the mechanic wrote a 500$ invoice. The customer did not understand how one THUMP could cost so much. The mechanic then itemized the invoice: one THUMP, 10$. Knowing where to THUMP, 490$.
2
... I think that a key aspect of expertise is having a large assortment of tools at our disposal, and knowing when to use which tool. Unfortunately, all the customer (and the novice data scientist) sees is us picking up a single tool and using it. The hard-won experience that allowed us to select the appropriate tool, or the thought processes that went into the selection, are invisible.
2
Something completely different: gold tag badge holders can unilaterally close a question as a duplicate. Sometimes I feel like this unilateral closure privilege should also apply to the "unclear what you are asking" closure reason (or, as it now reads more charitably, "needs more details"). I assume something like this would need to be done network-wise... any thoughts? Set up a meta question for discussion?
 
7 hours later…
14:25
@StephanKolassa I think documenting our thought process can be valuable, but I'll admit it is an area I could improve.
14:42
@StephanKolassa that anecdote? Well, thanks for now I am going to drop it at proper time.
14:58
The downvotes are strong with this one: meta.stackexchange.com/q/399619/256777
 
2 hours later…
17:17
@StephanKolassa Great presentation! youtube.com/watch?v=Q5LvodkPmSA
 
3 hours later…
20:03
I guess if you have a clean sinus wave then you don't need tissue waves either. 👃🤧

https://stats.stackexchange.com/questions/646635/calculate-the-probability-of-a-peak-in-cyclical-data-using-historical-observatio?noredirect=1#comment1210337_646635

« first day (5011 days earlier)      last day (292 days later) »