Figure 3 - available via license: Creative Commons Zero 1.0
Content may be subject to copyright.
Download
View publication
MSE results of order-α, isotonic CQR, and isotonic CER with n = 1000.

MSE results of order-α, isotonic CQR, and isotonic CER with n = 1000.

Source publication
Figure 3: MSE results of order-α, isotonic CQR, and isotonic CER with n...
Figure 5: MSE results of convexified order-α, CQR, and CER with n = 1000.
Figure 6: Bias results of convexified order-α, CQR, and CER with n = 1000.
Classification of nonparametric quantile-like estimators.
Percentage of expectile estimation's MSE less than quantile...
Partial frontiers are not quantiles
Preprint
Full-text available
  • May 2022
Quantile regression and partial frontier are two distinct approaches to nonparametric quantile frontier estimation. In this article, we demonstrate that partial frontiers are not quantiles. Both convex and nonconvex technologies are considered. To this end, we propose convexified order-$\alpha$ as an alternative to convex quantile regression (CQR)...
Cite
Download full-text

Contexts in source publication

Context 1
... consider the choice of quantiles τ . Figure 3 depicts the MSE results in estimating the quantile functions for different input dimensions and SNR specifications, while keeping the sample size fixed at n = 1000. In all scenarios considered, isotonic CQR and isotonic CER have far smaller MSE values than order-α. ...
Context 2
... Figures 3 and 5, we notice that the MSE statistic for each estimator decreases to a great extent once imposing the concavity constraint, especially for order-α. For instance, in the one-input case with σ = 1.88, the average MSE of convexified order-α for the five estimated quantiles decreases by more than 160% compared to its original counterpart. ...