|
Up Up With Tuples@bretthooverwx.bsky.social |
I feel like there is some value in recognizing where the mean track starts to cut across ensemble members rather than follow them. I wonder if there is a simple way to quantify this as a measurement of ensemble mean representativeness of its members.
2 replies 0 reposts 0 likes
|
Jeff Duda
@jeffdudaphdco.bsky.social
[ View ] |
Here's an idea: when it gets to that point, stop thinking of the mean as a consensus and instead remember that it's just an arithmetic mean, and the mean of a skewed data set (relevant if "survivor bias" is being considered) is usually not representative of most of the mass of the PDF. Go median.
1 replies 0 reposts 0 likes
|
James Franklin
@franklinjamesl.bsky.social
[ View ] |
Well, a simple count of the number of members in each time step would be a useful warning to watch out.
0 replies 0 reposts 0 likes