论文部分内容阅读
本文结合一个实例，讨论了多层次模型（ｍｕｌｔｉｌｅｖｅｌｍｏｄｅｌ）在分析重复观察数据中的应用。重复观测数据前后相关，且时间点不能随机安排，而本文的实例不仅有多个处理组，而且观测时间点也不等距，传统的统计分析方法难以胜任。多层次模型将时间、处理等作为变量或哑变量，通过模型的构造，将模型误差分解为重复测定间误差（一级误差）和受试动物间误差（二级误差），从而找出每个层次上影响误差构成的主要因素，并得到预测值随时间的总体变化趋势。
In this paper, an example is used to discuss the application of the multilevel model in the analysis of repeated observation data. Repeated observation data are related before and after, and the time points cannot be randomly arranged. However, the examples in this paper not only have multiple processing groups, but also the observation time points are not equidistant. The traditional statistical analysis methods are difficult to perform. The multi-level model takes time, processing, etc. as variables or dummy variables. Through the construction of the model, the model error is decomposed into repeated inter-assay errors (primary errors) and inter-animal errors (secondary errors) to find out each The main factors affecting the error composition at the level, and the overall trend of the predicted value over time.