Hi everyone,
I?ve noticed that some leaderboard submissions are showing extremely high hazard ratios (e.g., around 125), which seems quite unusual based on my experience with survival models and hazard ratios in biological studies. Typically, HR values tend to be much lower, even in strong predictive models.
I'm curious if anyone else has encountered similar results or could clarify whether such high values are expected given the nature of the data (methylation profiles) or if they might indicate potential issues.
Thanks in advance for any insights or guidance!
Created by Jakob T jtraeuble Thanks we are investigating and have determined we will need to change the statistical measure. We are working on selecting one that will be less gameable but still measure the important metrics we care about. Hey Jakob,
As the author of the submissions, I have so many possible answers in my head...
"All is fine. Nothing to see here. Please proceed..."
"You are right, there is an issue.
A SKILL issue!"
"The Dark Side is a pathway to many abilities some consider to be... unusual"
![The-dark-side-of-the-Force-is-a-pathway-to-many-abilities-some-consider-to-be-unnatural](https://www.magicalquote.com/wp-content/uploads/2016/09/The-dark-side-of-the-Force-is-a-pathway-to-many-abilities-some-consider-to-be-unnatural.jpg)
Sorry, could not resist ;-)
Seriously:
I can only recommend to take a break of a few days. The organizers are informed since last week, I assume there will be an update soon.
Drop files to upload
Unusually High Hazard Ratio in Phase 2 page is loading…