均值结构模型及其在医学中的应用

发布时间：2018-12-16 19:22

【摘要】： 医学、心理学及社会学研究中存在许多不可直接观测的现象,通常采用潜变量(latentvariable)对之进行描述。分析潜变量常采用结构方程模型(structure equation modeling,SEM),其中测量模型反映了可观测变量与潜变量之间的关系,结构模型反映了潜变量之间的结构关系,能够处理潜变量和测量误差是结构方程模型特有的优势。目前结构方程模型在医学领域的应用越来越多,通过考察方差/协方差结构,获得结构效度与关联性的结果。这样往往忽略了潜变量的均值结构与测量模型的截距。如果研究人员仅为评价潜变量之间的效应,以上分析尚可,但有时研究人员希望对比不同组的同一个或多个对应潜变量的均值与测量模型的截距,此时则须拟合均值结构模型(means structure model)。用结构方程分析比较潜变量的组间均值差异,相比传统的方差分析有以下的优点:(1)用潜变量可以正确调整测量上的误差,而传统的方差分析一般情况下不能够处理由信度所导致的问题;(2)未考察测量工具(如量表)的可比性而不合理地选用了方差分析,而用结构方程模型可进行测量的等同性检验(即可比性),在具有可比性的前提下,采用均值结构模型分析;(3)结构方程模型还能处理只有部分等同性的(partialinvariance)的多组比较,而方差分析却不能处理这类问题。本文系统详细地讨论了均值结构模型的统计原理,同时实现了均值结构模型的多组比较,并将均值结构模型应用于医学研究中。利用抑郁症病例——对照临床研究中的特质应对方式问卷(TCSQ)、社会支持(SSS)量表进行分组比较。验证性的均值结构模型结果显示患抑郁症的人消极应对方式高于正常人,而积极应对方式低于正常人;获主观支持、客观支持和对支持的利用都要比正常人低。均值结构方程模型结果进一步表明患抑郁症的人获主观支持、客观支持和对支持的利用都要低于正常人;并且在控制了主观支持与客观支持的效应的前提下,客观支持对支持的利用的作用无统计学意义,而对支持利用的作用主要来自于主观支持。
[Abstract]:There are many unobservable phenomena in medical, psychological and sociological research, which are usually described by the latent variable (latentvariable). The structural equation model (structure equation modeling,SEM) is often used to analyze latent variables, in which the measurement model reflects the relationship between observable variables and latent variables, and the structural model reflects the structural relations between latent variables. The ability to deal with latent variables and measurement errors is the unique advantage of the structural equation model. At present, the structural equation model is applied more and more in the field of medicine. By investigating the variance / covariance structure, the results of structural validity and correlation are obtained. In this way, the mean structure of the latent variable and the intercept of the measuring model are often ignored. If the researchers were only to evaluate the effects between the underlying variables, the above analysis would be fine, but sometimes the researchers wanted to compare the mean values of the same or more corresponding latent variables in different groups with the intercept of the measuring model. In this case, the mean value structure model (means structure model). Must be fitted. Compared with the traditional ANOVA, structural equation analysis has the following advantages: (1) the errors in measurement can be adjusted correctly by using latent variables. However, the traditional ANOVA can not deal with the problems caused by reliability in general. (2) without examining the comparability of measurement tools (such as the scale) and using ANOVA unreasonably, the equivalence test (i.e. comparability), which can be measured by using structural equation model, can be carried out under the premise of comparability. The mean value structure model is used. (3) the structural equation model can also deal with multiple comparisons of (partialinvariance) with only partial equivalence, but ANOVA can not deal with this kind of problems. In this paper, the statistical principle of the mean value structure model is discussed in detail. At the same time, the multi-group comparison of the mean value structure model is realized, and the mean value structure model is applied to the medical research. The trait coping style questionnaire (TCSQ), Social support (SSS) was used to compare depression case-control clinical study. The results of the confirmatory mean structure model showed that the negative coping style of depression patients was higher than that of normal people, but the positive coping style was lower than that of normal people, and the subjective support, objective support and utilization of support were lower than those of normal people. The results of the mean structural equation model further showed that the patients with depression received subjective support, objective support and the use of support are lower than the normal people; On the premise of controlling the effect of subjective support and objective support, the effect of objective support on the utilization of support has no statistical significance, but the role of support utilization mainly comes from subjective support.
【学位授予单位】：山西医科大学
【学位级别】：硕士
【学位授予年份】：2008
【分类号】：R311

【参考文献】