The supposed control of leaf loss of Common beech (Fagus sylvatica L.) by climatic factors. [Mathematical and statistical remarks concerning 'Control of the so called 'leaf loss' of European beech (Fagus sylvatica L.) by the weather' (Gruber,2004, AF

摘要

This paper discusses the serious shortcomings in Gruber's contribution entitled 'Control of the so-called 'leaf loss' of European beech (Fagus sylvatica) by the weather' [See AFJZ (2004) 175/4-5, 83-94]. The selection of predictor variables, their specific weighting and additive connection to multi-annual weather indices in this study are exclusively aimed at attaining high coefficients of determination (R-square) of the derived multiple linear regression models. In order to change a negative into a positive relationship with the dependent variable, Gruber calculated its reciprocal (1/x) instead of reflecting the predictor variable on the x-axis. This caused an unintentional and uncontrolled data transformation. In addition, restrictions for the application of multiple regression analyses (overfitting of the model, multicollinearity and autocorrelation of variables) as well as necessary testing criteria were ignored. With this approach, random effects cannot be distinguished from real influences which entails serious consequences for the practical interpretation of the results. The presented models are neither suitable for a better understanding of the crown condition of beech and its temporal variation nor do they provide valuable tools for retrospective analyses or predictions. Therefore, the conclusions and claims presented by Gruber are also invalid..