https://scholars.tari.gov.tw/handle/123456789/23057
Title: | APPLICABILITY OF DATA TRANSFORMATION IN REGRESSION-MODEL FOR STABILITY ANALYSIS | Authors: | LU, HY | Keywords: | DATA TRANSFORMATION;STABILITY;regression analysis | Issue Date: | Sep-1992 | Publisher: | Agricultural Association of China | Journal Issue: | 159 | Start page/Pages: | 29-40 | Source: | 中華農學會報 = Journal of the Agricultural Association of China | Abstract: | Linear regression model is useful to evaluate the adaptation of a quantitative character to the change of environments. Its proper use requires that : ( i ) genotype x environment interactions are adequately explained by a linear function of the environment ; and ( ii ) the residual mean squares from the regression are homogeneous across all genotypes. The transformation of data is a well -known technique for reducing error heterogeneity. From a practical standpoint, the data are usually difficultly interpreted on the transformed scale of measurement. Questions therefore need to be asked about the consequences of using or omitting a transformation when theoretical reasons establish a clear necessity. This study is to confirm to applicability of data transformation by analysing various set of trial data, if their residual mean squares from the linear regression are not homogeneous across all genotypes. Results showed that although positive correlation between with and without data transformation occurred, an unstable genotype might be wrongly regarded as stable because of the removal of heterogeneity among regressions. Thus the data transformation is not an optimal approach to the error homogeneity of regression analysis in plant stability. Removal of outliers may be an useful alternative in homogenizing the residual variance because no effects on the assessment of stability was found. |
URI: | https://scholars.tari.gov.tw/handle/123456789/23057 | ISSN: | 0578-1434 |
Appears in Collections: | SCI期刊 |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.