Die Bodenkultur - Journal for Land Management, Food and Environment
Comparison of some different statistical analysis to eliminate soil effects
Two widely used nearest neighbour analysis are compared to block analysis and a special form of covariance analysis.
With a simulation study the results of these different methods were evaluated. Levels of soil, error and factor effects varied in practical relevant sizes. A further distinction was made between the number of blocks and the number of variants per block.
Some relevant results of this study are given below:
- With high soil effects all alternatives to block analysis show more exact results then block analysis.
- In almost all situations covariance models are superior to other methods in concern to parameter estimation and hypothesis testing.
- Nearest neighbour methods are of little use for hypothesis testing.
Key words: Nearest neighbour analysis, analysis of covariance analysis, paramter estimation, hypothesis testing, simulation.