First Advisor

Holighi, Khalil Shafie

Document Type

Dissertation

Date Created

4-11-2016

Embargo Date

11-19-2016

Abstract

The asymmetric loss function is used in a situation where a positive error may be more serious than a negative error of the same magnitude or vice versa. One of the most commonly used asymmetric loss functions is the linex loss. The linex unbiased predictor has been developed and applied to real world applications. This study investigated how the linex unbiased prediction behaves when time series processes, AR(p), MA(q) and ARMA (p,q), parameters are unknown and being estimated, with different levels of variance, forecast step, shape parameter and series length. It started with deriving the predictor for each time series process, computing this predictor, and then discussing its properties. Empirical studies of the behavior of this predictor were investigated by using the Monte Carlo simulation. The results of this study showed that, a simpler time series model produced values that were closer to the condition of linex unbiasedness than a complex model. The condition of linex unbiasedness was affected by the variance but not the sign of the linex loss function shape parameter. For any time series model and any condition, as series length increased, the condition of linex unbiasedness values approached zero. When the time series parameters are unknown, the prediction is asymptotically linex unbiased.

Keywords

Autoregressive moving average, Linex loss, Unbiased prediction

Extent

238 pages

Local Identifiers

Yang_unco_0161D_10465.pdf

Rights Statement

Copyright is held by the author.

Share

COinS