A quantile regression approach to generating prediction intervals

Article Abstract:

A quantile regression approach to the formulation of predictive distributions and prediction intervals is presented as an alternative method. It was developed with the aim of settling the issue of whether to estimate prediction intervals through a theoretical approach, which assumes that the method is optical, or through an empirical approach. The quantile regression approach is described as a mixture of the empirical and theoretical procedures because it employes the empirical fit errors and generates forecast error quantile models which are functions of the lead time as proposed by the theoretically obtained variance expressions. The use of quantile regression to the empirical fit errors eschews the optimality assumption of the theoretical approach and the normality assumption.

Author: Bunn, Derek W., Taylor, James W.
Analysis, Usage, Prediction theory, Confidence intervals

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA


Regression metamodels for simulation with common random numbers: comparison of validation tests and confidence intervals

Article Abstract:

Two methods of linear regression analysis are compared using a Monte Carlo simulation experiment. Metamodels of Rao's test and Kleijnen's cross-validation procedure show that Rao's test is the preferred method under conditions of normal simulation responce since this procedure provides satisfactory probability of coverage and suitable half-length. Under lognormality, cross-validation gave more satisfactory type I errors based on estimates obtained through Ordinary Least Squares. Rao's test continues to provide satisfactory results in uniform distributions. The efficiency of the cross-validation method increases as higher correlations are generated by common seeds.

Author: Kleijnen, Jack P.
Models, Evaluation, Specifications

User Contributions:

Comment about this article or add new information about this topic:

CAPTCHA



Subjects list: Regression analysis
This website is not affiliated with document authors or copyright owners. This page is provided for informational purposes only. Unintentional errors are possible.