Evaluating the use of the automated unified structural equation model for daily diary data.
Evaluating the use of the automated unified structural equation model for daily diary data.
复制标题
DOI：
10.1080/00273171.2016.1265439
复制
发表时间：
2017
影响因子：
3.8
通讯作者：
Gates KM
中科院分区：
文献类型：
--
作者：
Lane ST;Gates KM
虽然有大量的资源描述了研究人员为什么以及如何收集时间序列数据，但很少有资源涉及心理科学中的时间序列数据分析（Hamaker，Ceulemans，Grasman，& Tuerlinckx，2015）。研究人员可以将个体之间的时间序列连接起来以得出一个模型，或者研究人员可以单独分析每个个体，不利用样本的共同信息。最近的研究表明，一种上级方法可能是建立个人层面的模型，但利用样本中共享的信息（Gates & Molenaar，2012）。组迭代多模型估计（GIMME）框架提供了一种使用统一SEM（Kim，Zhu，Chang，Bentler和Ernst，2007）估计由滞后效应和同期效应组成的个体水平结构方程模型的自动化程序（见图1）。在GIMME框架中，每个单独的模型都由组、（潜在的）子组和个人级别的路径组成。GIMME的性能已经在静息状态fMRI数据的背景下得到了很好的证明（Mumford & Ramsey，2014），其特征在于相当大的自回归（AR）效应。相比之下，每日日记数据的特征往往是小得多，潜在的非显着AR效应，特别是如果所研究的结构发生快于观察的时间尺度。目前尚不清楚是否纳入小的自回归效应将允许可靠的恢复滞后和同期的关系。进行了模拟研究，以调查GIMME在日常日记研究中遇到的条件下的性能。操作了各种模拟因素：变量数量（5，10）;时间点数量（30，60，90，120）;个体数量（25，75，150）;以及模型估计开始时存在滞后关系。同期关系设为β= 0.5，AR效应设为较小，β= 0.2。GIMME的表现
While there is a wealth of resources delineating both why and how researchers should collect time series data, fewer resources speak to the analysis of time series data within psychological science (Hamaker, Ceulemans, Grasman, & Tuerlinckx, 2015). Researchers may concatenate time series across individuals to arrive at one model, or researchers may analyze each individual separately, utilizing no information common to the sample. Recent research suggests that a superior approach may be to arrive at individual-level models but utilize information shared across the sample (Gates & Molenaar, 2012). The group iterative multiple model estimation (GIMME) framework presents an automated procedure for estimating individual-level structural equation models composed of both lagged and contemporaneous effects (see Figure 1) using the unified SEM (Kim, Zhu, Chang, Bentler, & Ernst, 2007). In the GIMME framework, each individual model is composed of group-,(potentially) subgroup-, and individual-level paths. The performance of GIMME has been well documented in the context of resting-state fMRI data (Mumford & Ramsey, 2014), which is characterized by sizable autoregressive (AR) effects. By contrast, daily diary data are often characterized by much smaller, potentially nonsignificant AR effects, particularly if the construct under study occurs faster than the time scale of observation. It is unknown whether the inclusion of small autoregressive effects will allow for the reliable recovery of the lagged and contemporaneous relationships. A simulation study was conducted to investigate the performance of GIMME under conditions encountered in daily diary research. A variety of simulation factors were manipulated: number of variables (5, 10); number of timepoints (30, 60, 90, 120); number of individuals (25, 75, 150); and the presence of the lagged relationships at the start of model estimation. Contemporaneous relationships were set to β= 0.5, and the AR effects were set to be small, β= 0.2. GIMME’s performance
影响因子：
5.7
作者：
Gates, Kathleen M.;Molenaar, Peter C. M.
通讯作者：
Molenaar, Peter C. M.
影响因子：
5.4
作者：
Hamaker, E. L.;Ceulemans, E.;Tuerlinckx, F.
通讯作者：
Tuerlinckx, F.
影响因子：
4.8
作者：
Kim, Jieun;Zhu, Wei;Ernst, Thomas
通讯作者：
Ernst, Thomas