| Abstrakt: |
This article proposes a regression-based stiffness test for pallets based on sample period reduction by trimming. The deflection data are collected from the stiffness tests for pallets, of which one group is from virgin pallets, produced by a Korean and a European vendor, and the other group is from recycle pallets made by a Korean vendor. The data consist of the test sample in normal as well as high temperature condition. Applying the cumulative sum of residuals (CUSUM) and the likelihood ratio (LR) methods, we identify the exact points of structural breaks, thereby rejecting the time-constancy of regression relationship. Specifically, the regression relationship turns out heavily influenced by the initial period of the tests. We remedy the potential biases stemming from the initial period by trimming the initial period out of the sample. It is shown that forecasting performances are substantially improved by the truncation of the first hour data. [ABSTRACT FROM AUTHOR] |
