中文摘要：

[目的]探讨如何利用结构方程模型研究变量平均水平随时间变化的趋势及变化过程中的个体间差异,为重复测量数据的分析提供方法学参考。[方法]以治疗急性脑梗死病为例,利用SAS中的CALIS过程,以线性与非线性潜生长曲线模型对102个病例的欧洲脑卒中量表评分时间上的变化情况进行分析。[结果]经过治疗后,欧洲脑卒中量表总分的线性与非线性潜生长曲线模型的拟合卡方值为43.411 9（P﹤0.000 1）与1.631 2（P=0.201 5）;在非线性模型中线性效应与二次方效应的统计指标均有统计学意义。[结论]非线性模型的拟合结果比线性模型好,且初始评分与治疗后的变化率均存在个体差异,变化率呈现非线性的速度增加,且变化率增加是随时间逐渐变小的;病例个体的初始总分越高,其变化率越大。

英文摘要：

[Objective] To explore the change trend of the mean level of variables -along with time and individual differenee during the changes by using structural equation modeling （SEM）, so as to provide methodological references for the analysis of repeated measured data. [ Methods] Taking the treatment of acute cerebral infarction as example, CALLS procedure of SAS software was used to fit linear and nonlinear latent growth curve model, which were used to analyze the change of total score of European stroke scale （ESS） of 120 eases along with time. [Results] The values of ehi square and P values of linear and nonlinear latent growth curve model were 43.411 9 （P 〈 0.000 1 ） and 1.631 2 （P = 0.201 5）, respectively. Linear effects and square effects in the nonlinear latent growth curve model both had statistical significance. [Conclusions] The nonlinear latent growth curve model is better than the linear model. The total score of ESS in initial level and the change rate of ESS along with time have individual variability. The increase of change rate of ESS is nonlinear and diminishes along with time. And the change rate of ESS is more rapid ff the total score of ESS in initial level is higher.