利用空间相关模型和传统方差分析模型对小麦和玉米3个田间试验产量数据进行了拟合与分析。结果表明，3个试验都显著存在空间变异，空间变异方差占据剩余变异方差的83.5%~0.4%；相对于传统随机完全区组分析法，空间相关模型法效应比较的标准误平均降低18.4%~14.2%，分析相对效率平均为1.50~1.36，因而比区组控制空间变异更有效；不同空间相关模型分析的结果呈现出一定的差异。建议利用空间相关模型分析田间试验，并利用Akaike信息准则（AIC）进行最佳空间相关模型选择。
Spatial variability often exists among field experimental units because of many factors, such as moisture, fertility, pH, and structure of soil, and the pressure of diseases and pests. Spatial variability can be dealt with in one of two ways: either though designs, by blocking to account for spatial effect, or though statistical adjustment, by nearest neighbor or trend analysis. Recently, models with spatial covariance structures such as those used in geostatistics have been proposed to account for spatial variability of field experiments. The objectives of this study were to investigate the spatial variation of field experiment, to compare the performance of spatial correlation models with classical variance analysis models, and to investigate the influence of spatial covariance structure selection on data analysis, based on fitting and analyzing yield data of 3 trails using wheat and corn varieties. The results showed that the spatial variation significantly existed in all trails; the spatial variation variance amounted to 83.5%–70.4% of the residual variation variance. Compared with classical randomized complete block analysis, the spatial correlation model analysis reduced standard error of effect contrast by 18.4%–14.2%, its average relative efficiency was 1.50–1.36, hence, it was more effective than the blocking in controlling the spatial variation. The results from different spatial correlation models were not identical. The spatial correlation model and selection of the optimal model using Akaike’s information criterion (AIC) are suggested for analysis of field experiments.
全 文 ：