---

作 者 ：胡希远 任常宏 吴冬

期 刊 ：植物营养与肥料学报 2011年 17卷 5期 页码：1205-1211

关键词：土壤养分；空间与时间；分析方法；线性混合模型；

Keywords:soil nutrient, space and time, Analysis method, mixed linear model,

摘 要 ：以土壤耕作方式对三叶草中氮素在土壤矿化影响试验为例，论述了涉及空间和时间效应的土壤养分试验的特点、可能的分析方法及线性混合模型分析的程序，并对方差分析与线性混合模型分析在该类试验统计测验结果的差异状况进行了比较。结果表明，该类试验的误差在不同深度土层间和时段间存在相关性和异质性, 其相关性随着土层间距和时段间距减小而增大。相对线性混合模型分析法，方差分析法分析该试验时，依测验的效应不同会导致第一类或第二类统计错误率的增大。线性混合模型分析法能考虑该类试验数据误差相关和异质的特性而对数据拟合效果好，所以利用其分析该类试验结论更可靠。

Abstract:The characteristics, possible analysis methods and programs for mixed model analysis of soil nutrient trials with spatial and temporal effects were studied by using data collected from a field experiment in which the effects of tillage on nitrogen mineralization of leguminous in soil was conducted. The difference between the analysis of variance and the mixed linear model in analysis of these trials was discussed. The results show that the correlations between the trial errors are depended on different soil layers and time periods. The correlations are increased with narrow spatial and temporal distances. Compared with the mixed linear model analysis, the analysis of variance could result in higher type I or II statistical error rates. The mixed linear model could deal with error correlation and heterogeneity and fit these trial data well, therefore, the analysis of these trials using the mixed linear model is more reliable.

---