0
The 0.95 Expectile Is Not the 95th Percentile
https://towardsdatascience.com/the-095-expectile-is-not-the-95th-percentile/(towardsdatascience.com)The 0.95 expectile and the 95th percentile are distinct statistical measures derived from different loss functions. Quantiles minimize asymmetrically weighted absolute errors, making them dependent on rank and coverage, while expectiles minimize asymmetrically weighted squared errors, making them sensitive to the magnitude of observations. For a uniform distribution, the 0.95 expectile covers only about 81% of the data, whereas the 95th percentile covers 95% by definition. The choice between using an expectile or a quantile should be determined by the specific cost function of the problem, as demonstrated through mathematical derivations and Python examples.
0 points•by hdt•5 hours ago
Comments (0)
No comments yet. Be the first to comment!
Have an account? Log in to join the discussion.